# Connector directory Source: https://docs.ideaboxai.com/connectors/connector-directory Browse all available IdeaBoxAI connectors across different categories and integration types. This directory provides a comprehensive view of all connectors available in IdeaBoxAI. Connectors enable seamless integration with external services, databases, and tools to power your knowledge bases and workflows. ## Accessing connectors Navigate to **Settings > Connections > Connectors** to view all available integration options. The Connectors page displays all available services you can connect to your IdeaBoxAI workspace. Connectors page showing all available integration options ## Finding the right connector Use the **search bar** at the top of the Connectors page to quickly find the connector you need. Type the name of the service or tool you want to connect, and the list will filter to show matching options based on your use case. Connectors page with search bar to find specific connectors for your use case ## Sample connectors Below are some sample connectors that are commonly used in IdeaBoxAI implementations. These connectors cover a wide range of use cases from data analytics to communication and productivity tools. Sample connectors showing Google Drive, Confluence, and other integration options Additional sample connectors including database and analytics platforms ## Available connector categories The following sections list all available connectors organized by category and use case. ### Data and analytics connectors Connect to SQL warehouses and data platforms to enable live data queries through knowledge bases. | Connector | Description | | -------------- | ---------------------------------------------------------------------------------------------------------------- | | **Snowflake** | Connect to a SQL warehouse through a knowledge base. Enables conversational AI and Agentic BI on Snowflake data. | | **PostgreSQL** | Direct connection through host, port, and database credentials. Works with any relational data. | | **MySQL** | Direct connection through host, port, and database credentials. Works with any relational data. | | **ClickHouse** | Direct connection for analytical workloads. Optimised for high-performance queries. | | **SQL Server** | Connect to Microsoft SQL Server databases with standard credentials. | | **Actian** | Direct database connection for Actian data warehouses. Enables pipeline analytics and reporting. | | **Zen** | Direct connection for Zen database systems. | ### CRM and sales connectors Integrate with customer relationship management platforms and sales intelligence tools. | Connector | Description | | -------------------------- | -------------------------------------------------------------------------------------------------------- | | **Salesforce** | Access leads, opportunities, and accounts. Pipeline overview, deal updates, and contact history. | | **HubSpot CRM** | Manage contacts, deals, and pipelines. Contact enrichment, deal stage progression, and activity logging. | | **Zoho CRM** | Access leads, accounts, and sales data. Deal management, contact timelines, and pipeline updates. | | **Microsoft Dynamics 365** | Connect to Dynamics 365 CRM and ERP data. Opportunity and contact management. | | **Gong** | Access call recordings and revenue intelligence. Call transcription and conversation analysis. | | **ZoomInfo** | Access company and contact intelligence. Enrich accounts with firmographic and technographic data. | ### Communication and productivity connectors Connect to email, messaging, and collaboration platforms. | Connector | Description | | ------------------- | ------------------------------------------------------------------------------------------- | | **Google Drive** | Find and access documents. Save AI-generated content to Drive folders. | | **Gmail (SMTP)** | Draft and send emails from a configured Gmail account. All sends appear in the Sent folder. | | **Google Calendar** | View upcoming meetings and access calendar data. | | **MS Outlook** | Access Outlook to send email, list inbox, and review messages. | | **MS Teams** | Access Teams channels, messages, and communications. | | **Slack** | Post messages, alerts, and summaries to Slack channels. | | **Confluence** | Search internal documentation and knowledge bases. Access pages and spaces. | ### Data catalog and metadata connectors | Connector | Description | | ----------------------- | ---------------------------------------------------------------------------------------- | | **Zeenea** | Connect to data catalog platforms. Pull metadata, lineage, and data quality information. | | **Zoho Book Analytics** | Connect to Zoho Books for financial and accounting analytics. | ## Next steps Set up database and document connections for your knowledge bases. Resolve common connection issues. # Knowledge base connections Source: https://docs.ideaboxai.com/connectors/knowledge-base-connections Learn how to connect databases, documents, and cloud warehouses to IdeaBoxAI knowledge bases for AI-powered querying. A knowledge base connection links IdeaBoxAI to your data. Whether that is a live relational database, a cloud data warehouse, or a library of documents, this guide covers all three types with step-by-step configuration. ## Knowledge base types IdeaBoxAI supports three knowledge base types. Each type is optimized for a different data format and query pattern. | KB type | Best for | Supported sources | | --------------------- | ---------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------- | | **Unstructured data** | Documents, PDFs, and text. Optimised for semantic search and AI reasoning. | Device upload, Google Drive, Confluence, Markdown and text. | | **Structured data** | Relational databases and spreadsheets. Supports SQL queries, cubes, analytics, and dashboards. | MySQL, PostgreSQL, Snowflake, ClickHouse, SQL Server, Actian, Zen, Zoho Book Analytics, CSV, and Excel. | ## Set up an unstructured data knowledge base Use this type for documents, playbooks, runbooks, and any content intended for semantic search. In the IdeaBoxAI sidebar, click **Knowledge Bases**. The dashboard shows all existing knowledge bases with name, status, type, creator, and last modified date. Knowledge Bases dashboard showing a searchable table of knowledge bases with Name, Status, Type, Created By, and Last Modified columns. Click **+ Create Knowledge Base**. A type selection modal appears. Select **Unstructured Data** and click **Next**. The Name and Description modal appears. Fill in the following fields: * **Name** (required): Use a clear naming convention, for example "Sales Playbooks 2026" or "RFP Answer Library". * **Description**: Explain what data this knowledge base contains and which personas will use it. * **Tags**: Add domain or team tags, for example "sales", "onboarding", "CRE". Click **Create**. Name and Description modal for creating an Unstructured Knowledge Base with fields for Name, Description, and Tags. The empty Data Source panel appears. Click **+ Add Data** to reveal four options. Add Data dropdown showing four options: Upload From Device, Add Markdown/Text, Upload From Google Drive, and Upload From Confluence. * **Upload From Device**: Drop files or browse to select them. Supports PDF, PNG, JPG, DOCX, TXT, MP3, WAV, OGG, and MP4. Batch upload is supported. * **Add Markdown/Text**: Paste or type content directly into the editor. Useful for SOPs, playbooks, or knowledge articles without a file. * **Upload From Google Drive**: Authenticate with your Google account, browse your Drive folders, and select files or folders to import. * **Upload From Confluence**: Authenticate with your Confluence workspace, browse spaces and pages, and select content to import. The following screenshot shows the Upload From Device modal with its drag-and-drop area. Upload From Device modal with a drag-and-drop area showing supported file formats: PDF, PNG, JPG, DOCX, TXT, MP3, WAV, OGG, and MP4. After upload, each file appears in the data source table with columns for file name, status, file type, and size. The status transitions from **Processing** to **Processed**. Data source table showing an uploaded PDF file with Processing status, file type, and size columns. Wait for all files to show **Processed** before connecting this knowledge base to an agent or automation. If a file shows a failure status, remove it and re-upload. Knowledge Bases dashboard showing a knowledge base with Processed status, confirming the data is ready. Ensure documents are machine-readable text PDFs (not scanned images) for best processing results. Re-index data after any bulk upload. ## Set up a structured data knowledge base Use this type for relational databases, spreadsheets, or any structured data source. Click **+ Create Knowledge Base**. Select **Structured Data**. Enter a name, description, and tags in the Name and Description modal, then click **Create**. Name and Description modal for creating a Structured Knowledge Base with fields for Name, Description, and Tags. The Connect Your External Knowledge Base page appears. Choose from two categories: **Spreadsheet**: CSV, Excel, or Google Sheets *(coming soon)*. **Database Connection**: MySQL, Zen, ClickHouse, SQL Server, PostgreSQL, Actian, Zoho Book Analytics, or Snowflake. Connect Your External Knowledge Base page showing spreadsheet options (CSV, Excel, Google Sheets) and database connection options (MySQL, Zen, ClickHouse, SQL Server, PostgreSQL, Actian, Zoho Book Analytics, Snowflake). If you selected a database connection, the Enter Database Credentials form appears. Fill in the following fields: * **Host** (required): Your database server address or IP. * **Port** (required): Database port (PostgreSQL: 5432, MySQL: 3306, Snowflake: 443). * **Username** (required): A database user with read access. * **Password** (required): The database password. Click **Test** to verify the connection. Enter Database Credentials form for MySQL showing required fields for Host, Port, Username, and Password with a Test button. If you selected CSV, the Upload CSV Files page appears. Click **+ Upload Files** to select CSV files from your device. Wait for all files to show **Processed**. Upload CSV Files page showing a processed CSV file with file name, status, type, and size columns. After a successful connection, IdeaBoxAI imports your database schema. The Datasets view shows: * **Left sidebar**: All database tables (Data Sets). * **Center panel**: Table relationships, joins, and column details. * **Options**: Enhance with AI, Add Semantics, and Define Relationships. Browse the tables and confirm the schema looks correct. Use **Define Relationships** to specify table joins. Cubes sit on top of your raw tables and define metrics (measures) and groupings (dimensions) for AI queries. Navigate to the **Cubes** tab and click **+ Add Cube** to open the Add New Cube modal. **Option A, AI Generate** (recommended for most cases): Enter a cube name, cube description, natural language query describing the analytics you want, and business context. Click **Generate**. The AI writes the SQL, measures, and dimensions. Add New Cube modal with AI Generate selected, showing fields for Cube Name, Cube Description, Natural Language Query, and Business Context. **Option B, Manual Query** (for advanced users): Enter a cube name, write custom SQL in the SQL Query editor, then define measures and dimensions manually using **+ Add Measure** and **+ Add Dimension**. Add New Cube modal with Manual Query selected, showing a SQL Query editor and sections for adding Measures and Dimensions. Click **Save Cube**. In the cube configuration panel, select measures, dimensions, and filters, then click **Run Query**. Results appear in the Results tab. Also check the **Generated SQL** tab to inspect the query and the **REST API** tab for programmatic access. Validate results against your source data. Navigate to **Agents**, select your agent, open the **Configuration** tab, select this knowledge base from the **Knowledge Base** dropdown, and click **Save**. The agent can now query your structured data using natural language. Always test the generated SQL in the cube query panel before connecting the knowledge base to a live agent. Start with simple cubes and iterate. ## Next steps Browse all available connectors by category. Build analytical models with measures, dimensions, and queryable APIs. Resolve common connection issues for databases and imports. # What are connectors? Source: https://docs.ideaboxai.com/connectors/overview Learn how IdeaBoxAI connectors bridge the platform to your external data sources, business tools, and communication services. Connectors in IdeaBoxAI bridge the platform to your external data sources, business tools, and communication services. Every piece of live data that an agent queries, every email it sends, and every CRM record it reads flows through a configured connector. Connectors are configured per-organization and scoped by persona permissions. Users only access data they are authorized to use. ## Accessing connections Navigate to **Connections** in the left sidebar. The Connections page is divided into two tabs, each covering a different integration category. The Connections page showing the Connectors tab with Google Drive, Confluence, Zeenea, and other service integrations | Tab | What it contains | | -------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Connectors** | OAuth and API-based service integrations โ€” Google Drive, Confluence, Zeenea, Salesforce, HubSpot, Zoho CRM, Dynamics 365, Gong, and ZoomInfo. Click **Connect** to authenticate each service. Use **+ Add Custom** to register a custom integration. | | **Databases** | Org-wide database connections (PostgreSQL, MySQL, Snowflake, and others). Named connections here are reusable across knowledge bases. Admin permissions may be required to add or remove connections. | ## How to connect a connector The following steps walk through connecting an external service to IdeaBoxAI. Google Drive is used as an example, but the same flow applies to all OAuth-based connectors. Go to **Settings** in the left sidebar to go to Connections. Profile section in the top corner of the platform The Connectors page displays all available integrations across different categories. Use the **search bar** to find a specific connector. For example, type "Google Drive" to filter the list and locate the Google Drive connector. Connectors page showing all available connector cards Click the **Connect** button on the Google Drive connector card. The platform navigates to the **Connect your Google account** screen. Sign in with your Google credentials and grant the requested permissions. Connect your Google account screen showing the Google OAuth sign-in prompt After authenticating, the Google Drive connector shows a **Connected** status on the Connectors page. The integration is now active and ready to use across agents, knowledge bases, and automations. Google Drive connector card showing Connected status The same steps apply to all OAuth-based connectors. For connectors that use API keys or tokens instead of OAuth, you will be prompted to enter your credentials in a modal after clicking Connect. ## Connection layers IdeaBoxAI uses connections to integrate with external data sources and services. Connectors are configured inside knowledge bases to provide access to databases, documents, and cloud services. | Layer | What it does | | ------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Knowledge base connections** | Database and document connectors configured inside a knowledge base. These provide access to structured or unstructured data, such as a PostgreSQL database, a Snowflake warehouse, or a set of uploaded documents. | ## How connectors power the platform IdeaBoxAI workflows depend on connectors to access data and services. The following table shows how connectors support different platform features. | Platform feature | Connection used | Example | | ----------------- | ------------------------------------------------------- | --------------------------------------------------------------------------------- | | Knowledge bases | Database connector (PostgreSQL, Snowflake, MySQL) | Query sales tables through an ODBC driver. | | Data queries | Structured KB connection | Query a data warehouse for pipeline statistics. | | Document search | Unstructured KB (Confluence, Google Drive, file upload) | Search Confluence for RFP answer library content. | | Email integration | Gmail or Microsoft Mail 365 connector | Send follow-up emails from a configured account. | | CRM integration | Salesforce, HubSpot, or Zoho CRM connector | Log call notes back into a Salesforce opportunity. | | Automations | All of the above, chained in sequence | Scheduled automation queries a knowledge base, generates a report, and emails it. | ## Core value proposition Connectors deliver measurable improvements by removing manual data retrieval and enabling AI-powered workflows. * Query live databases and documents through natural language without writing SQL or switching tools. * Send emails, post to Slack, and update CRM records directly from the agent chat interface. * Combine multiple data sources in a single agent for comprehensive, cross-system intelligence. * Automate multi-step workflows that span databases, documents, and communication tools. * Maintain security through per-persona permission scoping and organization-level configuration. ## Next steps Explore the following guides to configure your connectors. Connect databases, documents, and cloud warehouses to your knowledge bases. Browse all available connectors by category with setup requirements. Resolve common connection issues for databases and imports. # Toolkit directory Source: https://docs.ideaboxai.com/connectors/toolkits Browse toolkits. Click any toolkit to view its tools and triggers.

Categories

Showing 982 of 982 toolkits

9 abilities
5 abilities
11 abilities
3 abilities
6 abilities
4 abilities
14 abilities
8 abilities
21 abilities
162 abilities
17 abilities
2 abilities
3 abilities
8 abilities
1 abilities
39 abilities
20 abilities
99 abilities
4 abilities
3 abilities
40 abilities
18 abilities
12 abilities
40 abilities
2 abilities
23 abilities
1 abilities
7 abilities
0 abilities
30 abilities
3 abilities
20 abilities
9 abilities
5 abilities
15 abilities
16 abilities
1 abilities
2 abilities
16 abilities
18 abilities
4 abilities
4 abilities
29 abilities
9 abilities
6 abilities
4 abilities
18 abilities
31 abilities
4 abilities
28 abilities
5 abilities
4 abilities
11 abilities
31 abilities
0 abilities
11 abilities
0 abilities
6 abilities
8 abilities
90 abilities
5 abilities
0 abilities
2 abilities
2 abilities
3 abilities
11 abilities
0 abilities
4 abilities
4 abilities
31 abilities
159 abilities
8 abilities
18 abilities
0 abilities
12 abilities
2 abilities
18 abilities
2 abilities
21 abilities
11 abilities
5 abilities
1 abilities
36 abilities
9 abilities
8 abilities
22 abilities
26 abilities
5 abilities
35 abilities
17 abilities
3 abilities
6 abilities
3 abilities
17 abilities
18 abilities
0 abilities
0 abilities
314 abilities
20 abilities
10 abilities
0 abilities
15 abilities
1 abilities
7 abilities
4 abilities
6 abilities
14 abilities
153 abilities
0 abilities
8 abilities
10 abilities
9 abilities
9 abilities
293 abilities
0 abilities
6 abilities
0 abilities
18 abilities
19 abilities
35 abilities
10 abilities
0 abilities
6 abilities
9 abilities
18 abilities
10 abilities
7 abilities
17 abilities
0 abilities
4 abilities
25 abilities
33 abilities
3 abilities
12 abilities
20 abilities
5 abilities
10 abilities
141 abilities
13 abilities
42 abilities
2 abilities
27 abilities
17 abilities
5 abilities
14 abilities
25 abilities
32 abilities
93 abilities
12 abilities
7 abilities
20 abilities
8 abilities
1 abilities
4 abilities
8 abilities
43 abilities
6 abilities
9 abilities
11 abilities
1 abilities
8 abilities
0 abilities
4 abilities
9 abilities
15 abilities
32 abilities
21 abilities
5 abilities
36 abilities
183 abilities
126 abilities
42 abilities
19 abilities
6 abilities
34 abilities
8 abilities
16 abilities
24 abilities
4 abilities
34 abilities
9 abilities
4 abilities
4 abilities
105 abilities
18 abilities
5 abilities
23 abilities
22 abilities
21 abilities
1 abilities
0 abilities
11 abilities
6 abilities
41 abilities
6 abilities
30 abilities
13 abilities
80 abilities
15 abilities
3 abilities
3 abilities
0 abilities
48 abilities
10 abilities
11 abilities
18 abilities
10 abilities
2 abilities
23 abilities
0 abilities
3 abilities
28 abilities
14 abilities
3 abilities
6 abilities
5 abilities
5 abilities
5 abilities
14 abilities
11 abilities
6 abilities
1 abilities
45 abilities
42 abilities
10 abilities
6 abilities
16 abilities
22 abilities
7 abilities
447 abilities
42 abilities
0 abilities
0 abilities
9 abilities
0 abilities
9 abilities
5 abilities
4 abilities
13 abilities
0 abilities
61 abilities
10 abilities
10 abilities
0 abilities
25 abilities
192 abilities
1 abilities
14 abilities
6 abilities
47 abilities
7 abilities
164 abilities
19 abilities
15 abilities
26 abilities
0 abilities
13 abilities
4 abilities
10 abilities
9 abilities
12 abilities
11 abilities
7 abilities
17 abilities
5 abilities
4 abilities
2 abilities
10 abilities
342 abilities
26 abilities
12 abilities
29 abilities
11 abilities
18 abilities
7 abilities
1 abilities
4 abilities
10 abilities
2 abilities
11 abilities
0 abilities
3 abilities
17 abilities
8 abilities
16 abilities
27 abilities
13 abilities
1 abilities
5 abilities
0 abilities
3 abilities
57 abilities
125 abilities
13 abilities
5 abilities
2 abilities
7 abilities
21 abilities
14 abilities
26 abilities
22 abilities
8 abilities
2 abilities
2 abilities
4 abilities
0 abilities
52 abilities
7 abilities
2 abilities
14 abilities
28 abilities
3 abilities
95 abilities
3 abilities
4 abilities
24 abilities
9 abilities
14 abilities
49 abilities
7 abilities
5 abilities
3 abilities
41 abilities
1 abilities
9 abilities
117 abilities
0 abilities
19 abilities
16 abilities
27 abilities
18 abilities
49 abilities
7 abilities
3 abilities
10 abilities
6 abilities
10 abilities
15 abilities
14 abilities
123 abilities
7 abilities
11 abilities
15 abilities
3 abilities
6 abilities
1 abilities
14 abilities
24 abilities
23 abilities
45 abilities
12 abilities
99 abilities
1 abilities
20 abilities
31 abilities
1 abilities
24 abilities
5 abilities
5 abilities
8 abilities
0 abilities
2 abilities
7 abilities
0 abilities
4 abilities
21 abilities
9 abilities
7 abilities
9 abilities
5 abilities
6 abilities
3 abilities
17 abilities
14 abilities
2 abilities
3 abilities
24 abilities
9 abilities
29 abilities
22 abilities
13 abilities
19 abilities
827 abilities
54 abilities
18 abilities
7 abilities
29 abilities
4 abilities
25 abilities
4 abilities
53 abilities
5 abilities
9 abilities
2 abilities
14 abilities
4 abilities
31 abilities
23 abilities
6 abilities
6 abilities
5 abilities
1 abilities
35 abilities
42 abilities
57 abilities
9 abilities
13 abilities
52 abilities
7 abilities
269 abilities
19 abilities
32 abilities
43 abilities
19 abilities
8 abilities
17 abilities
32 abilities
5 abilities
16 abilities
7 abilities
47 abilities
6 abilities
0 abilities
48 abilities
24 abilities
58 abilities
18 abilities
0 abilities
7 abilities
3 abilities
15 abilities
36 abilities
29 abilities
35 abilities
16 abilities
13 abilities
17 abilities
27 abilities
6 abilities
24 abilities
39 abilities
16 abilities
3 abilities
122 abilities
255 abilities
143 abilities
3 abilities
4 abilities
21 abilities
3 abilities
25 abilities
5 abilities
1 abilities
7 abilities
16 abilities
11 abilities
32 abilities
26 abilities
1 abilities
1 abilities
2 abilities
44 abilities
112 abilities
9 abilities
10 abilities
3 abilities
16 abilities
47 abilities
10 abilities
51 abilities
0 abilities
8 abilities
4 abilities
1 abilities
2 abilities
33 abilities
7 abilities
10 abilities
4 abilities
49 abilities
21 abilities
6 abilities
2 abilities
6 abilities
27 abilities
20 abilities
17 abilities
24 abilities
6 abilities
49 abilities
1 abilities
29 abilities
231 abilities
8 abilities
11 abilities
10 abilities
0 abilities
15 abilities
6 abilities
4 abilities
4 abilities
2 abilities
13 abilities
7 abilities
10 abilities
24 abilities
7 abilities
9 abilities
4 abilities
9 abilities
1 abilities
9 abilities
0 abilities
30 abilities
18 abilities
4 abilities
0 abilities
2 abilities
0 abilities
24 abilities
1 abilities
4 abilities
5 abilities
17 abilities
4 abilities
4 abilities
26 abilities
1 abilities
10 abilities
7 abilities
13 abilities
7 abilities
2 abilities
4 abilities
12 abilities
24 abilities
12 abilities
11 abilities
1 abilities
1 abilities
275 abilities
56 abilities
6 abilities
38 abilities
23 abilities
3 abilities
1 abilities
49 abilities
11 abilities
3 abilities
13 abilities
4 abilities
10 abilities
6 abilities
4 abilities
5 abilities
43 abilities
8 abilities
7 abilities
3 abilities
16 abilities
266 abilities
1 abilities
0 abilities
2 abilities
1 abilities
28 abilities
0 abilities
14 abilities
8 abilities
0 abilities
96 abilities
21 abilities
0 abilities
25 abilities
15 abilities
21 abilities
21 abilities
5 abilities
15 abilities
0 abilities
1 abilities
0 abilities
5 abilities
5 abilities
33 abilities
13 abilities
7 abilities
11 abilities
12 abilities
6 abilities
16 abilities
69 abilities
12 abilities
0 abilities
1 abilities
24 abilities
12 abilities
9 abilities
0 abilities
43 abilities
9 abilities
9 abilities
2 abilities
27 abilities
9 abilities
21 abilities
40 abilities
48 abilities
0 abilities
22 abilities
29 abilities
0 abilities
4 abilities
1 abilities
17 abilities
8 abilities
10 abilities
44 abilities
28 abilities
2 abilities
8 abilities
7 abilities
24 abilities
14 abilities
5 abilities
4 abilities
8 abilities
7 abilities
17 abilities
2 abilities
48 abilities
3 abilities
5 abilities
353 abilities
14 abilities
5 abilities
0 abilities
5 abilities
3 abilities
1 abilities
2 abilities
10 abilities
24 abilities
3 abilities
3 abilities
13 abilities
4 abilities
112 abilities
1 abilities
1 abilities
33 abilities
1 abilities
19 abilities
15 abilities
0 abilities
1 abilities
0 abilities
69 abilities
8 abilities
19 abilities
0 abilities
5 abilities
0 abilities
30 abilities
13 abilities
278 abilities
15 abilities
4 abilities
3 abilities
22 abilities
15 abilities
7 abilities
1 abilities
1 abilities
6 abilities
8 abilities
18 abilities
14 abilities
11 abilities
24 abilities
22 abilities
0 abilities
358 abilities
0 abilities
146 abilities
0 abilities
1 abilities
2 abilities
20 abilities
13 abilities
67 abilities
5 abilities
11 abilities
0 abilities
8 abilities
3 abilities
4 abilities
3 abilities
4 abilities
6 abilities
19 abilities
0 abilities
32 abilities
13 abilities
24 abilities
0 abilities
1 abilities
31 abilities
16 abilities
22 abilities
27 abilities
2 abilities
6 abilities
8 abilities
20 abilities
32 abilities
10 abilities
35 abilities
10 abilities
5 abilities
3 abilities
2 abilities
0 abilities
0 abilities
130 abilities
38 abilities
8 abilities
21 abilities
11 abilities
18 abilities
8 abilities
6 abilities
10 abilities
15 abilities
11 abilities
11 abilities
3 abilities
15 abilities
26 abilities
7 abilities
0 abilities
19 abilities
33 abilities
3 abilities
3 abilities
2 abilities
12 abilities
13 abilities
6 abilities
0 abilities
70 abilities
0 abilities
6 abilities
1 abilities
5 abilities
7 abilities
41 abilities
27 abilities
19 abilities
10 abilities
0 abilities
5 abilities
5 abilities
0 abilities
1 abilities
5 abilities
10 abilities
12 abilities
16 abilities
3 abilities
14 abilities
36 abilities
37 abilities
7 abilities
9 abilities
12 abilities
375 abilities
2 abilities
5 abilities
10 abilities
1 abilities
176 abilities
0 abilities
23 abilities
1 abilities
18 abilities
0 abilities
17 abilities
28 abilities
5 abilities
0 abilities
6 abilities
13 abilities
0 abilities
95 abilities
28 abilities
18 abilities
2 abilities
122 abilities
0 abilities
2 abilities
0 abilities
2 abilities
9 abilities
4 abilities
8 abilities
1 abilities
19 abilities
3 abilities
51 abilities
19 abilities
2 abilities
3 abilities
13 abilities
114 abilities
114 abilities
16 abilities
2 abilities
19 abilities
32 abilities
5 abilities
15 abilities
6 abilities
22 abilities
6 abilities
8 abilities
18 abilities
8 abilities
22 abilities
21 abilities
91 abilities
2 abilities
0 abilities
3 abilities
22 abilities
8 abilities
6 abilities
9 abilities
11 abilities
3 abilities
2 abilities
41 abilities
4 abilities
10 abilities
0 abilities
4 abilities
40 abilities
57 abilities
79 abilities
8 abilities
17 abilities
10 abilities
33 abilities
6 abilities
8 abilities
37 abilities
49 abilities
1 abilities
11 abilities
5 abilities
1 abilities
16 abilities
10 abilities
129 abilities
1 abilities
18 abilities
4 abilities
17 abilities
28 abilities
4 abilities
11 abilities
2 abilities
1 abilities
7 abilities
21 abilities
13 abilities
33 abilities
6 abilities
12 abilities
13 abilities
10 abilities
9 abilities
1 abilities
13 abilities
1 abilities
2 abilities
6 abilities
1 abilities
6 abilities
32 abilities
31 abilities
5 abilities
10 abilities
19 abilities
0 abilities
1 abilities
327 abilities
3 abilities
0 abilities
2 abilities
15 abilities
12 abilities
3 abilities
22 abilities
72 abilities
1 abilities
35 abilities
5 abilities
3 abilities
1 abilities
36 abilities
33 abilities
4 abilities
12 abilities
18 abilities
18 abilities
2 abilities
3 abilities
42 abilities
12 abilities
5 abilities
35 abilities
7 abilities
2 abilities
2 abilities
16 abilities
0 abilities
5 abilities
1 abilities
17 abilities
5 abilities
1 abilities
28 abilities
21 abilities
10 abilities
4 abilities
1 abilities
20 abilities
16 abilities
18 abilities
19 abilities
5 abilities
10 abilities
0 abilities
5 abilities
0 abilities
0 abilities
21 abilities
38 abilities
3 abilities
14 abilities
6 abilities
38 abilities
13 abilities
8 abilities
4 abilities
8 abilities
4 abilities
0 abilities
0 abilities
1 abilities
18 abilities
14 abilities
16 abilities
11 abilities
43 abilities
0 abilities
18 abilities
7 abilities
6 abilities
12 abilities
0 abilities
0 abilities
0 abilities
6 abilities
0 abilities
28 abilities
17 abilities
111 abilities
3 abilities
8 abilities
# Troubleshooting connections Source: https://docs.ideaboxai.com/connectors/troubleshooting Resolve common connection issues for IdeaBoxAI knowledge base connectors and data imports. This guide covers the most common connection issues in IdeaBoxAI and how to resolve them. If a problem persists after following these steps, contact the platform team. ## Structured knowledge base issues The following table covers issues related to database connections in structured data knowledge bases. | Problem | Resolution | | ------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | **Connection refused** | Check that the host and port are correct and reachable. Verify firewall rules allow inbound connections from the IdeaBoxAI platform. | | **Authentication failed** | Confirm the username and password are correct. Use a read-only service account. For SSL issues, confirm whether the database requires SSL and toggle the **SSL Mode** setting accordingly. | | **Schema import hangs** | The database may have too many tables. Filter to relevant schemas before connecting if possible. Contact the platform team if the issue persists. | | **Cube query returns no data** | Check that the selected measures and dimensions are valid for the cube. Inspect the **Generated SQL** tab for errors and ensure the underlying tables are not empty. | ## Unstructured knowledge base issues The following table covers issues related to document uploads and external source imports. | Problem | Resolution | | ----------------------------- | ------------------------------------------------------------------------------------------------------------------------------------- | | **File stuck in Processing** | Remove the file and re-upload. Ensure the PDF is machine-readable (not a scanned image). For large files, split into smaller batches. | | **Confluence import fails** | Regenerate the Atlassian API token. Confirm the user has read access to the selected spaces. Try importing a smaller space first. | | **Google Drive import fails** | Re-authenticate with Google (OAuth token may have expired). Check that the user has access to the selected folders. | ## Next steps Review setup steps for database and document connections. Browse all available connectors by category. # Set up full integration for a sales persona Source: https://docs.ideaboxai.com/connectors/tutorials/sales-integration Learn how to configure all required connections for a Sales Account Executive persona on IdeaBoxAI. This tutorial walks through configuring all required connections for a Sales Account Executive (AE) persona on IdeaBoxAI. At the end, the Sales AE agent will query live pipeline data, search Confluence for playbooks, research prospects on the web, and send emails. All of this happens from a single chat interface. The IdeaBoxAI Co-pilot home screen showing an Account Executive persona with suggested skill prompts The home screen shows a personalized landing page for the Sales AE persona. Skill prompts are auto-detected based on the agent's configuration. As you complete the setup below, these prompts will be backed by live data. ## Target architecture The following table shows the connections you will configure and the purpose each one serves. | Connection | Purpose | | ---------------------------------- | ------------------------------------------------------------------------------------- | | **Structured KB (data warehouse)** | Live pipeline data, deal analytics, quota tracking, and forecast views. | | **Unstructured KB (Confluence)** | Sales playbooks, battlecards, RFP answers, objection library, and onboarding content. | | **Unstructured KB (Google Drive)** | Proposal templates, case study documents, pricing files, and email templates. | | **Gmail tool** | Draft and send follow-up emails, outreach sequences, and meeting invites. | | **Web Search tool** | Company profiles, news, job postings, and tech stack signals for prospect research. | ## Phase A: Create the structured knowledge base (pipeline data) Navigate to **Knowledge Bases > + Create Knowledge Base > Structured Data**. Enter a name (for example, "Sales Data Warehouse"), a description (for example, "Live pipeline, deal, account, and quota data"), and relevant tags. Click **Create**. Enter your database connection credentials: host, port, database name, username (read-only service account), and password. Enable **SSL Mode** if required. Click **Connect** and wait for schema import to complete. Click **+ Add Cube > AI Generate** and create cubes for your key analytics views. For example: * **Pipeline Overview**: "Total open deals by stage, value, and close date." Business context: "B2B SaaS sales pipeline tracking." * **Deal Health Monitor**: "Deals with no activity in 5+ days and high value." Business context: "Sales deal risk monitoring." * **Quota Attainment**: "Closed won value compared to quota target by rep for the current quarter." Business context: "Sales quota tracking." Click **Generate** for each cube, then **Save**. Run test queries on each cube and validate results match your source data. ## Phase B: Create the unstructured knowledge base (Confluence sales content) Navigate to **Knowledge Bases > + Create Knowledge Base > Unstructured Data**. Name it (for example, "Sales Knowledge: Confluence"), add a description and tags, then click **Create**. Click **Add Data > Upload from Confluence**. Authenticate with your Atlassian email and API token. Select the spaces to import: for example, Sales Playbooks, Competitive Intelligence, RFP Library, and Onboarding and Training. Click **Import** and monitor status until all pages show **Processed**. ## Phase C: Create the unstructured knowledge base (Google Drive templates) Navigate to **Knowledge Bases > + Create Knowledge Base > Unstructured Data**. Name it (for example, "Sales Templates: Google Drive"), add a description and tags, then click **Create**. Click **Add Data > Upload from Google Drive**. Authenticate with Google, navigate to your sales folders (for example, Proposal Templates, Case Studies, Pricing, Email Sequences), and select them. Click **Import** and wait for all files to show **Processed**. ## Phase D: Create and configure the Sales AE agent Navigate to **Agents > + Create Agent**. In the General tab, enter a name (for example, "Sales AE Agent"), a description, and relevant tags. Set visibility to **Workspace** and click **Save**. Open the **Configuration** tab and navigate to the **Instructions** box. Either write manually or use AI-assisted generation. Include the agent's role, capabilities, and any behavioral guidelines. In the **Configuration** tab, scroll to the **Knowledge Bases** dropdown. Select all three knowledge bases you created: the structured data warehouse, the Confluence knowledge base, and the Google Drive knowledge base. Click **Save**. Navigate to the **Tools** tab. Add and configure the following: * **Gmail (Google SMTP)**: Enter the sending email address, App Password, SMTP host, and port. * **Tavily Web Search**: Enter the API key. Click **Save**. If you use a CRM platform, navigate to **Connections > CRM** and connect your CRM before finishing. The CRM tab lists all supported platforms โ€” Salesforce, HubSpot, Zoho CRM, Dynamics 365, Gong, and ZoomInfo โ€” and shows how many are currently connected. The CRM tab in Connections showing Salesforce, HubSpot, Zoho CRM, Dynamics 365, Gong, and ZoomInfo with Connect buttons Click **Connect** next to your platform and follow the OAuth flow. Once connected, the Sales AE agent can read pipeline data, log call notes, and update deal stages directly from the chat interface. Open the Sales AE Agent chat and test each connection with these prompts: * **Pipeline data**: "How is my pipeline looking this quarter? Show me open deals by stage." * **Confluence**: "What does the playbook say about handling security objections?" * **Google Drive**: "Show me the structure of the proposal template." * **Web search**: "What is the latest news about \[prospect company name]?" * **Email**: "Draft a follow-up email to John at Acme Corp after our call today." Verify each response pulls from the correct data source. Always ask the agent for confirmation before sending any email. Review the drafted content and recipient list before approving. ## Next steps Explore the full sales persona setup and walkthrough scenarios. Browse all available connectors by category. Resolve common connection issues. # Admin Console Source: https://docs.ideaboxai.com/copilot/admin-console Manage users, personas, connections, and security policies from the Admin Console. The Admin Console is accessible to users with the **Admin** persona. It provides controls over the entire workspace, including user management, persona management, connections, and security policies. All changes made in the Admin Console affect the entire workspace. Only users assigned the **Admin** persona can access the Admin Console. ## User Management Manage all users in your workspace and control their access. The user management table displays the following columns. | Column | Description | | ----------------- | ------------------------------------------------------------------------------ | | **User Details** | Avatar, full name, and email address. | | **Groups** | Groups the user belongs to. | | **Roles** | Role pills showing assigned roles (e.g. System Admin Role, Demand Admin Role). | | **Status** | Active or inactive toggle. | | **Last Modified** | Timestamp of the most recent change. | | **Actions** | Edit and manage user settings. | Use the **+ Add User** button to invite new users to the workspace. Use the search bar to quickly find users by name or email. User Management table showing users with their groups, roles, status, and last modified dates. ## Persona Management Create, edit, and manage workspace personas. The persona management table displays the following columns. | Column | Description | | ----------------- | ------------------------------------------- | | **Persona** | Name of the persona. | | **Description** | Short description of the persona's purpose. | | **Users** | Number of users assigned to the persona. | | **Status** | Active or inactive status. | | **Last Modified** | Timestamp of the most recent change. | Use the **+ Create New Persona** button to add a new persona to the workspace. Each persona defines a distinct role with its own skills, knowledge bases, and data source access. Persona Management table showing personas with descriptions, user counts, status, and last modified dates. ## Connection management Admin-level connection management determines which integrations are visible to users. Only approved connections appear in user **Settings > Connections**. The header displays the total count of approved connections. ### Connection list The connection list is presented as a table with the following columns. | Column | Description | | -------------- | ------------------------------------------------------------------------------------------ | | **Connection** | Icon, name, and short description of the integration. | | **Category** | Badge indicating the type: `API`, `MCP`, or `DATABASE`. | | **Status** | Current connection status (e.g. `Not connected`, or saved connection count). | | **Actions** | **Connect** button to initiate setup, or **Settings** to reconfigure existing connections. | | **Approved** | On/Off toggle controlling user visibility. | The list is filterable by **All**, **MCP**, **APIs**, and **Databases**, and includes a search bar for quick lookup. **Approved ON** means users can see and connect the integration in their own Settings. **Approved OFF** hides the integration from all non-admin users. Admin connection management table showing all integrations with approved toggles, connection status, category badges, and action buttons. ### APIs Filter by **APIs** to view all API-based integrations. Each row shows the connection name, category badge, connection status, and an **Approved** toggle. Use the **Connect** button to initiate OAuth or API key setup, and **Settings** to reconfigure existing connections. Admin Connections page filtered by APIs, showing API integrations with status, actions, and approved toggles. ### Databases Filter by **Databases** to view database providers. Each database connection shows the provider name, category badge, saved connection count, and an **Add connection** action for configuring additional database instances. Admin Connections page filtered by Databases, showing database providers with saved connection counts and add connection actions. ## Security Granular access control organized into four tabs: **Resources**, **Policies**, **Roles**, and **Groups**. ### Resources tab Resources are named groups that represent collections of platform entities. The resources table displays the following columns. | Column | Description | | ----------------- | --------------------------------------------------------------------------- | | **Name** | The resource group name. | | **Entity** | Badge indicating the type: Dashboard, Knowledge Base, Project, or Resource. | | **Last Modified** | Timestamp of the most recent change. | #### Add resource modal Choose from the available entity types: Dashboard, Knowledge Base, Project, User, Resource, Policy, Role, or Group. Enter a **Resource Name** and **Description**. Build conditional rules using a field dropdown, operator, and value. Combine conditions with **AND** / **OR** logic. Security Resources tab showing named resource groups with entity badges and last modified dates. ### Policies tab Policies define access rules by combining a **subject** (role), **actions** (permissions), and a **resource**. The policies table displays the following columns. | Column | Description | | ----------------- | ------------------------------------------------------------------------------------ | | **Policy Admin** | Name and description of the policy. | | **Resources** | The resource group the policy applies to. | | **Actions** | Permission icons indicating granted actions: Create, Converse, Update, View, Delete. | | **Roles** | Roles assigned to this policy. | | **Last Modified** | Timestamp of the most recent change. | Use the **+ Add Policy** button to create a new access rule. Security Policies tab showing policy list with role assignments and last modified dates. ### Roles tab Roles are named permission levels that control what users can do within the workspace. The roles table displays the following columns. | Column | Description | | ----------------- | ----------------------------------------------------------------------- | | **Role Name** | Name and description of the role. | | **Users** | User pills showing assigned members, with a count for additional users. | | **Policies** | Policy pills showing attached policies. | | **Groups** | Associated groups. | | **Last Modified** | Timestamp of the most recent change. | Use the **+ Add Role** button to create a new role. Use the **Filters** button to narrow down the list. Security Roles tab showing default roles with user pills, policy counts, and group counts. ### Groups tab Groups are named collections of users for bulk role assignment. The groups table displays the following columns. | Column | Description | | ----------------- | -------------------------------------------------------------------- | | **Name** | Group name. | | **Users** | User pills showing group members, with a count for additional users. | | **Roles** | Role pills showing assigned roles. | | **Last Modified** | Timestamp of the most recent change. | Use the **+ Add Group** button to create a new group. Use the **Filters** button to narrow down the list. Security Groups tab showing group list with user pills, roles, and last modified dates. # Build dashboards with the Persona AI Source: https://docs.ideaboxai.com/copilot/artifacts/dashboards Give the Persona AI your data and describe the outcome you want, and it builds an interactive dashboard artifact with the right charts and layout. Dashboards are a special, interactive artifact type. You give the Persona AI some data and describe what you want to see, and it builds the dashboard for you, no manual chart-building required. This guide covers providing data, creating a dashboard, refining it, and the limitations to keep in mind. ## Provide the data A dashboard is built from data you give the Persona AI. You can provide it in three ways. * **Upload a file** - CSV, Excel, PDF, and other common formats. * **Paste data** - Paste it directly into the chat. * **Use a connected source** - Where your workspace has one set up. The Persona AI reads what you provide and works from that. If your file has multiple sheets or sections, tell the Persona AI which part matters. ## Create a dashboard With data in hand, describe the outcome you want in plain language, not the chart mechanics. For example: > "Here's our usage export. Build a dashboard showing adoption over time, the top Skills, and which teams are most active." The Persona AI then does the following. 1. Reads your data and identifies the relevant metrics. 2. Chooses appropriate visualizations, for example, line, bar, table, and KPI cards. 3. Lays them out into a single dashboard artifact in the side panel. ## Refine a dashboard Iterate conversationally to shape the dashboard. * **Add or remove views** - "Add a card for total active users." / "Drop the geography map." * **Change visualizations** - "Make the usage trend a bar chart instead." * **Filter and segment** - "Break this down by team." / "Only show the last 30 days." * **Reorder layout** - "Move the KPI cards to the top." ## Good practices A few habits produce a sharper dashboard. * **Give clean, well-labeled data.** Clear column headers and consistent formatting help the Persona AI pick the right metrics and charts. * **Be specific about the question.** "How is adoption trending?" produces a sharper dashboard than "show me some metrics." * **Name your time window and segments.** Specify the period, for example, "this quarter", and how you want data sliced, for example, by team, by Skill, or by org. * **Download or link a snapshot** when you need to share results, since a downloaded dashboard is a point-in-time snapshot rather than a live view. ## Limitations to keep in mind Dashboards built this way are fast and conversational, with a few trade-offs. * A dashboard reflects the data you provided when it was created. To update it, give the Persona AI fresh data, as it doesn't refresh on its own. * A shared or downloaded dashboard is a snapshot from when it was created, not a live, continuously updating view. * Very large files or highly custom analytics may need a dedicated analytics tool. The Persona AI is built for fast, conversational dashboards rather than deep BI workflows. For deep, continuously updating business intelligence, see the [Agentic BI](/agentic-bi/introduction) tab. # Create artifacts Source: https://docs.ideaboxai.com/copilot/artifacts/overview Learn how the Persona AI creates standalone artifacts in a side panel, and how to edit, iterate on, keep, and reuse them. Artifacts are standalone pieces of content that the Persona AI creates for you inside a conversation, documents, dashboards, diagrams, app interfaces, code, and more. Instead of being buried in the back-and-forth of a chat, an artifact opens in its own panel beside the conversation, where you can view it, edit it, iterate on it, and share it. This guide covers how artifacts get created and how to edit, keep, and reuse them. ## How artifacts get created You don't need a special command to create an artifact. When you ask the Persona AI for something substantial enough to stand on its own, it generates an artifact automatically and opens it in the side panel next to your chat. Typical prompts that produce an artifact include the following. * "Write a one-page product brief for our onboarding redesign." * "Build a dashboard showing weekly active users by team." * "Create a flowchart of our approval workflow." * "Draft a landing page for the new Skills marketplace." The conversation stays on the left, and the artifact renders on the right. As you keep chatting, the Persona AI updates the same artifact rather than starting over. ## Edit and iterate Once an artifact exists, you refine it by asking in chat. Describe the change in plain language, for example, "make the headline shorter", "add a column for last login", or "switch the chart to a bar graph", and the Persona AI revises the artifact in place. Keep iterating in the same conversation until it's right. ## Keep and reuse Artifacts live in the conversation they were created in, so you can return to that chat anytime to view or continue iterating. To use an artifact elsewhere, you have two options. * **Download it** to your device in a compatible format. * **Share a link** that others can open. See [Share artifacts](/copilot/artifacts/sharing). If you'll need an artifact later, download it or grab a share link before moving on. Artifacts stay tied to their original conversation, so they're easy to lose track of otherwise. # Share artifacts Source: https://docs.ideaboxai.com/copilot/artifacts/sharing Share an artifact with an expiring link or download it in a compatible format for use outside IdeaBoxAI. When an artifact is ready to reach other people, you can share it with a link or download it. This page covers both options and how expiring links keep shared content from lingering. ## Share a link Generate a shareable link to the artifact and send it to whoever needs it. Anyone with the link can open the artifact in their browser, no IdeaBoxAI account required. Share links expire. When you create a share link, you set how long it stays active. Once that window passes, the link stops working, and the artifact is no longer accessible through it. This keeps shared content from lingering longer than you intend. Choose a shorter expiry for anything sensitive, and a longer one for links you expect people to return to. If a link expires and you still need to share, generate a new one. ## Download For use outside the platform, download the artifact in a compatible format, for example, documents as PDF or markdown, tables as CSV, and visuals as images. Available formats depend on the artifact type. A downloaded file is a snapshot. Later changes to the artifact won't update what you've already downloaded. # Types of artifacts Source: https://docs.ideaboxai.com/copilot/artifacts/types See the artifact types the Persona AI supports, from documents and dashboards to diagrams, tables, web pages, code, and charts. IdeaBoxAI supports several artifact types. The Persona AI picks the most appropriate type based on your request, but you can always ask for a specific one. This page lists the available types and a few things worth knowing about how they behave. ## Available types The table below shows each type, what it's best for, and an example request. | Type | Best for | Example request | | ------------------------- | -------------------------------------------- | -------------------------------------------------- | | **Document** | Briefs, specs, reports, written content | "Write a PRD for the Community Hub." | | **Dashboard** | Metrics, KPIs, data views | "Show signups and activation rate this quarter." | | **Diagram / Flowchart** | Processes, system maps, user flows | "Diagram the asset-sharing permission flow." | | **Table / Data view** | Structured lists, comparisons, trackers | "Make a feature comparison table vs. competitors." | | **Web page / UI** | Landing pages, prototypes, interface mockups | "Prototype a settings screen for AI Agents." | | **Code** | Scripts, components, technical snippets | "Write a script to clean this CSV." | | **Chart / Visualization** | Single-focus data visuals | "Plot revenue by month as a line chart." | ## Things worth knowing A few behaviours apply across all artifact types. * **One artifact, one focus.** Each artifact is a single deliverable. If you ask for "a report and a dashboard", you typically get two separate artifacts. * **Interactive artifacts.** Dashboards, UI prototypes, and some charts are interactive. You can filter, toggle, and click within them, not just view them. * **Switching types.** You can ask the Persona AI to convert content from one type to another, for example, "turn this table into a chart". Dashboards are a special, interactive artifact type. See [Build dashboards with the Persona AI](/copilot/artifacts/dashboards) for how to create one from your data. # Create a persona with AI Source: https://docs.ideaboxai.com/copilot/create-persona-ai Build a ready-to-use persona in seconds by answering a few questions about its role, focus, and tone, then review and add it to your workspace. Instead of writing out all the instructions for a persona yourself, **Create with AI** asks you a few simple questions: the role, what it focuses on, and its tone, and builds a ready-to-use persona for you in seconds. This guide walks through each step, from opening Persona Management to adding the finished persona to your workspace. ## Open Persona Management Start from the Persona Management page, where every persona in your workspace is listed. Click **Personas** in the left menu. The **Persona Management** page shows all the personas in your workspace. In the top-right corner, click **+ Create persona**. A menu appears with two choices: * **Create with form** - Fill in every detail yourself. * **Create with AI** - Answer a few questions and let it build the persona for you. Select **Create with AI** to open the guided flow. Persona Management page showing the Create persona dropdown with the Create with AI option. ## Choose the role The **Create Persona with AI** window opens. Across the top, you see four steps: **Role**, **Focus**, **Tone**, and **Preview**. In the **Role** step, you are asked, "What role should this persona represent?" Type or choose the job this persona does for your team. For example: * Sales Development Rep * Customer Support Agent * Financial Analyst Make your choice, or type your own, then move to the next step. Role step in the Create Persona with AI dialog showing the role question and example roles. ## Choose the focus Next, you are asked, "What domain and data should it work with?" Pick one or more areas that describe what this persona should know about and work with. The options include: * **Pipeline and CRM data** * **Budgets and spend** * **Product usage and metrics** * **Contracts and policies** You can also type your own focus in the text box, for example, "Pipeline health from our Salesforce CRM". This tells the persona which information to lean on when it answers your questions. Focus step showing domain and data options such as Pipeline and CRM data, Budgets and spend, Product usage and metrics, and Contracts and policies. ## Choose the tone Then you are asked, "What communication style should it use?" Pick the tone that fits how you want this persona to sound. The options include: * **Formal** * **Concise and direct** * **Warm and consultative** * **Data-driven** You can also type your own, for example, "Concise and data-driven". Tone step showing communication style options such as Formal, Concise and direct, Warm and consultative, and Data-driven. ## Review your persona Once you answer all three questions, your persona is built. The **Preview** step shows a loading screen while it is being created. Generating persona loading state in the Preview step. When it is ready, the **Preview** step shows the following. * **Name** - A name created from your answers, for example, *Budget-Focused Sales Rep*. You can change it here. * **Avatar color** - The color for the persona's icon. Click any color to change it. * **Instructions** - A full set of instructions describing the persona's role, what it does, and how it behaves. Read these carefully, as they shape how the persona answers in every chat. * **Suggested skills** - A few skills recommended for this persona, for example, *budget-allocation-optimizer*, *spend-pattern-analyzer*, and *sales-target-achievement-tracker*. These are already ticked, so untick any you don't want. Preview step showing the generated persona name, avatar colour, instructions, and suggested skills. If the persona isn't quite what you had in mind, click **Regenerate** to create a new version from the same answers. You can also change the name and the instructions right here before you save. ## Add it to your workspace When you're happy with the preview, click **Add to workspace**. The persona is created and appears on the **Persona Management** page. It is switched on and ready to use straight away. After creating a persona with AI, open it and adjust the **instructions** to match how your team actually works and the words you use. Learn how to share it in [Share a persona](/copilot/share-persona). # Create a skill with the Persona AI Source: https://docs.ideaboxai.com/copilot/create-skill-ai Describe the skill you want in plain language and let the IdeaboxAI Persona AI plan, write, and package a finished skill file you can install. The quickest way to make a skill is to tell the **IdeaboxAI Persona AI** what you want in your own words. It plans the skill, checks with you to make sure it understood, writes all the instructions, and gives you a finished skill file you can install. This guide walks through describing, reviewing, building, and installing a skill. ## Open the Persona AI and describe your skill Start from the Persona AI home screen and describe what you need. Click **Home** in the left menu to open the **IdeaboxAI Persona AI**. In the chat box, describe the skill you'd like. Try to be clear about three things: * What the skill should **do**. * When you'd want to **use** it. * What its finished **result** should look like. For example, "Can you create a skill for market and competition research?" Persona AI home screen with the skill creation prompt typed in the chat input. ## Look over the plan Before writing anything, the Persona AI shows you a plan so you can check it first. It shows the following. * **Skill name** - A suggested name, for example, `market-competition-research`. * **What it would do** - A short list of what the skill covers. * **When you'd use it** - The kinds of questions or situations it's meant for. * **What the result looks like** - The sections the finished output will have, for example, a summary, a market overview, competitor profiles, a comparison, opportunities and risks, and sources. The Persona AI might also ask you a couple of quick questions, for example, whether the skill should be general or focused on one industry, whether it should do a quick check or a deep dive, and how you'd like the result laid out. Answering these questions gives you a skill that fits your needs better. If you're happy with a solid general version, just reply "go ahead". Persona AI showing the skill plan with What it would do, When it would trigger, and Output format sections. ## The Persona AI builds your skill Once you say yes, the Persona AI writes the whole skill for you: the instructions, the layout for the results, a guide for choosing good sources, and any extra supporting files. You see it working as it goes. When it's finished, you see the following. * A **skill card** with the skill's name and version, for example, `market-competition-research` Skill V1. * A **Download** button to save the skill file. * A **Share** button to send the skill to someone. * A short recap of what the skill does, what's included, and when to use it. * **Next steps** for how to install and try out your new skill. Persona AI showing the completed skill card with Download and Share buttons, and a full summary. ## Install your skill With the skill file saved, upload it to Skills Studio to make it available to your personas. Click **Download** to save the skill file to your computer. Go to **Skills Studio** by clicking **Skills** in the left menu. Click **+ Create Skill**, then **Upload**, and choose the file you just saved. Your skill now appears in Skills Studio, ready to add to your personas. ## Tips for creating good skills A little upfront detail produces a sharper skill on the first try. * **Be clear about the result you want.** The more you describe how the finished output should look, its sections, layout, and length, the more useful your skill will be. * **Say when you'd use it.** Tell the Persona AI what kinds of questions this skill should help with. This helps it know when to step in. * **Answer the questions it asks.** These questions clear up anything unclear, and answering them gives you a better skill on the first try. * **Try it before adding it to a persona.** Once installed, run three to five realistic questions through it to check the results before you put it to work in a live persona. * **Keep improving it with the Persona AI.** If the result isn't quite right, tell the Persona AI what you'd like to change in the same chat, and it makes a new, improved version. * **Keep each skill focused.** A skill that does one thing well is easier to look after and works more reliably than one that tries to do too much at once. Once your skill is installed, add it to a persona from the **Edit Persona** panel. See [About skills](/copilot/skills-overview) for how skills connect to personas, and [Share a skill](/copilot/share-skill) to send it to teammates. # Getting started with the Persona AI Source: https://docs.ideaboxai.com/copilot/getting-started Navigate the home screen, switch between personas, use the chat input, and access the Admin Console. This guide walks you through the Persona AI interface. By the end, you will know how to navigate the home screen, switch between personas, use skill cards, access conversation history, and open the Admin Console. ## Home screen The home screen is your primary workspace. It displays a personalized greeting, skill cards tailored to the active persona, and a persistent chat input bar. Conversational AI home screen showing the personalised greeting, skill cards in a 2-column grid, and chat input bar. ### Greeting The home screen displays a personalized welcome message along with a description of the Persona AI's capabilities. > **Good morning, \[name]!** > Describe what you need below; skills are auto-detected, or type `/` to pick one. ### Quick starter prompts Six quick-start prompts are displayed in a **2-column grid**, tailored to the active persona. Close-up of the six quick starter prompts in a 2-column grid, each with an icon and a natural language prompt. * **Icon and prompt**: Each card contains an icon paired with a natural language prompt that represents a common task for the current persona. * **One-click pre-fill**: Clicking a card pre-fills the chat input with that prompt, so you can review it before sending. * **Persona-tailored**: Cards change when you switch personas, reflecting the tasks and prompts relevant to each role. ### Chat input bar The chat input bar is **persistent at the bottom** of the home screen. * **Placeholder text:** *Ask the Persona AI anything... (type / for skills)* * **Attach button:** The **+** icon on the left allows you to attach files or context to your message. * **Submit button:** The arrow icon on the right sends your message. * **Status indicator:** Displays **Ready to help** (green dot) in the top bar, or a processing state while the system is working. The chat input remains visible on every screen so you can send a message at any time without navigating back to the home screen. ## Persona switching Each persona surfaces different skill cards and knowledge base scope. Switching personas updates all of these immediately. Click the persona selector in the **top-right corner** of the header. Choose the desired persona from the list. A checkmark indicates the currently active persona. You can also click **+ Create New** to create a persona directly from this dropdown. The home screen refreshes: skill cards and knowledge base scope all reflect the newly selected persona. Persona switcher dropdown in the top-right corner showing available personas with a checkmark on the active persona and a Create New option. Switching personas does not affect your conversation history. You can change personas mid-session and still access previous conversations. ## Chat History panel Click the **history icon** (clock) in the **top-left header** to expand the Chat History panel as a left sidebar. Chat History panel expanded on the left showing past conversations grouped by recency. * **Grouped by recency:** Conversations are organized under headings such as *Last 7 Days*. * **Truncated summaries:** Each entry shows a short summary of the conversation topic with a relative timestamp. * **New conversation:** Click the **+** button at the top of the panel to start a fresh conversation. * **Close the panel:** Click the **x** button to collapse it. * **Resume with full context:** Click any history entry to reopen that conversation with its complete context intact. ## Accessing the Admin Console Access User Management, Connections, Security, Observability, and Audit Logs from the Admin Console. Only users with admin privileges can access this area. The Admin Console's left navigation opens; there, you can access User Management, Connections, Security, Observability, and Audit Logs. Profile dropdown menu in the bottom-left corner showing Dark mode toggle, Admin Console, Settings, and Sign Out options. # What is the Persona AI? Source: https://docs.ideaboxai.com/copilot/introduction Learn how the IdeaBoxAI Persona AI uses skills and connected data to deliver role-specific intelligence, content generation, and task execution. The IdeaBoxAI Persona AI is a role-aware, AI-powered assistant embedded within the IdeaBoxAI platform. It uses purpose-built skills to handle research, content generation, analytics, and task execution, acting as an intelligent layer on top of your organization's data and knowledge systems. The Persona AI adapts its behavior based on the active persona. Capabilities are available across all roles, but the depth, context, and outputs adjust automatically. ## Core value proposition The Persona AI delivers measurable improvements across your organization. * Reduce research and information gathering from hours to minutes. * Auto-generate documents, emails, and briefs from structured templates and live data. * Surface actionable insights and intelligence proactively from connected data sources. * Scale best practices and standardized processes across teams. * Enable teams to work more efficiently with AI-assisted task execution and content generation. ## Core capabilities The Persona AI provides the following core capabilities. The depth, context, and specific outputs vary by persona configuration. * **Research and information retrieval**: Answers research questions by querying connected data sources and the web. Pulls company profiles, financial filings, news, job postings, and tech stack signals. Responses include source citations. * **Analytics and intelligence**: Answers analytical questions about key metrics and performance indicators. Generates summary statistics, trend analysis, rankings, and forecast views. * **Day-to-day task assistance**: General-purpose AI assistant for daily work. Summarizes documents, drafts communications, explains concepts, researches topics, and prepares for meetings. * **Artifact and document generation**: Generates downloadable DOCX, PDF, and structured emails. Combines templates with live data from connected sources. * **Skills-based task execution**: Each persona is equipped with purpose-built skills that handle specific tasks for that role. For example, a Project Manager persona has skills for portfolio health scanning, owner report generation, and cash flow forecasting. A Sales Engineer persona has skills for competitive battle cards and POC scoping. Skills define what the persona can do and are triggered automatically based on your request. * **Conversation history and continuity**: All conversations are saved and accessible from the History panel. Resume any past conversation with full context retained. ## How it works The Persona AI uses intent classification to process every request. The Persona AI receives the input and classifies it by intent: **research**, **draft**, **analyze**, **coach**, or **automate**. Based on the classified intent and the active persona, the Persona AI selects the appropriate skill and passes structured context to it. The skill queries the relevant data sources, processes the information, and returns a grounded response to the user. Skill selection is **persona-aware**. The active persona determines which skills are available and what default context is included with each request. ## Data source and knowledge access The Persona AI pulls information from connected data sources in real time to inform answers, generate artifacts, and populate outputs. Data access is gated by the active persona's permissions and the connections configured in Settings. Users only see data they are authorized to access. The following table describes the supported data sources and the information each one provides. | Data Source | What it provides | | ------------------ | ------------------------------------------------------------------------------------------------------------------ | | **Confluence** | Queries pages, spaces, and articles to surface internal documentation, knowledge articles, and onboarding content. | | **Data Warehouse** | Direct database connection. Query structured data for performance analytics, reporting, and forecast views. | | **Google Drive** | Retrieve templates, documents, and other unstructured content from connected folders. | | **Salesforce** | CRM integration. Live account data, contact records, activity history, and opportunity details. | | **Gong** | Call summaries, transcript highlights, and conversation signals for meeting prep and analysis. | | **Gmail** | Draft and send emails and meeting invites. Requires explicit user confirmation before sending. | New data source integrations can be added from **Settings > Connections**. Each integration is configured per-organization and scoped by persona permissions. ## Next steps Explore the following guides to get started with the Persona AI. Navigate the home screen, switch personas, and access the Admin Console. Create personas, configure permissions, and assign users. # Meet your Persona Source: https://docs.ideaboxai.com/copilot/meet-your-persona Your AI colleague who already knows your data, your role, and your business context. Not a chatbot. Not generic AI. A team member built around how you actually work. Most AI tools start from zero every time. You type a question, get a generic answer, and then spend the next twenty minutes correcting it with context the tool should already have. Your job codes, your vendors, your pipeline, your KPIs, your approval workflows. Every conversation is a cold start. A Persona is different. It is an AI colleague that already understands your role, your data, and the specific way your business operates. It knows that cost code 4100 is "Structural Steel" on Job 2401. It knows that NovaTech's renewal is stalled at the procurement stage. It knows your AP threshold is \$25,000 for auto-approval. And it remembers what you asked yesterday. **Your data, your role, your AI, working together from day one.** ## What your Persona knows A Persona is not a single feature. It is the combination of three layers that produce role-specific intelligence no generic AI tool can replicate. Purpose-built capabilities for your role. An AP Coordinator gets invoice triage and lien deadline tracking. A Sales Engineer gets competitive battle cards and POC scoping. Each skill is designed for a specific workflow, not a generic prompt. Your documents, policies, and guides. Coverage manuals, RFP libraries, underwriting rules, product documentation. The Persona queries these in real time and cites the specific section in its response. Live connections to your business systems. Salesforce opportunities, ERP job costs, Snowflake risk tables, Google Drive templates. The Persona queries structured data directly and returns current numbers, not cached summaries. These three layers combine into something no standalone tool offers. When a Project Manager asks "which jobs are over budget?", the Persona does not guess. It runs a live query against your connected ERP system, cross-references the commitment log, identifies the five cost codes burning fastest, and recommends specific next steps. When a Sales Engineer asks "draft an RFP response for this requirement", the Persona pulls from your RFP library, competitive battlecards, and product documentation to generate a tailored response. The Skills define *what* it can do. The Knowledge Base defines *what it knows*. The Data Sources define *what it can see right now*. ## See the difference The gap between a generic AI answer and a Persona answer is not subtle. It is the difference between a suggestion and a decision-ready briefing. ### "Which accounts need attention this week?" "To identify which accounts need attention, consider reviewing accounts with upcoming renewals, those with recent support tickets, and any accounts where engagement has dropped. You might want to check your CRM for accounts with low activity scores or upcoming contract dates. Prioritize based on revenue impact and relationship health." "3 accounts need attention this week: **NovaTech (ARR \$340K):** Contract renewal in 18 days. Last activity was 12 days ago (demo follow-up email, no reply). Single-threaded through procurement. Risk: high. Recommended action: schedule executive sponsor intro before renewal window closes. **Meridian Health (ARR \$180K):** Open support ticket #4891 (P1, unresolved 6 days). CSAT dropped from 4.2 to 3.1 in Q4. Risk: medium. Recommended action: loop in CS lead before QBR on Friday. **Vertex Financial (ARR \$95K):** Expansion opportunity flagged. They downloaded the enterprise pricing page twice this week, and the champion mentioned 'scaling to the compliance team' in the last Gong call. Risk: none. Recommended action: send expansion proposal by Thursday." The generic answer tells you *how to think about the question*. The Persona answer tells you *what to do today*, backed by live CRM data, call transcripts, and activity signals. Every account name, every dollar amount, every recommendation comes from your connected systems. ### "What is the lien deadline exposure this week?" "Lien deadlines vary by state and contract type. Generally, you should track preliminary notice deadlines (typically 20-30 days from first furnishing), mechanic's lien filing deadlines (60-90 days from last furnishing), and payment bond claim deadlines. Check your state's specific requirements and review your open invoices against these timelines." "2 invoices have lien deadline exposure this week: * **INV-2847 (Pacific Concrete, \$42,300):** Lien deadline is Monday, April 28. Payment is approved but not yet released. The vendor's lien waiver for the prior draw is on file. Recommended action: release payment today to clear before deadline. * **INV-2891 (Summit Electric, \$18,750):** Lien deadline is Wednesday, April 30. Invoice is in dispute; the billed amount exceeds the committed amount on PO-1847 by \$3,200. Recommended action: contact Summit Electric to reconcile the overage before releasing partial payment." A generic AI gives you a textbook overview of lien law. The Persona gives you the two invoices that will cause legal exposure this week, the exact dollar amounts from your AP system, and the specific action to take on each. ## Memory that builds over time Most AI tools forget everything the moment you close the tab. Every session is a cold start. You re-explain your projects, re-state your preferences, re-provide context that the tool already had yesterday. A Persona remembers. When you tell the Persona that you prefer revenue reports broken out by product line rather than region, it remembers that for every future report. When you explain that "Project Cascade" refers to Job 2401 internally, it never asks again. When you spend a session building a competitive analysis, the next conversation picks up exactly where you left off. This conversational memory creates compounding value. The more you use the Persona, the more context it retains, and the less time you spend repeating yourself. After a month, it understands your shorthand, your priorities, and the way you like information presented. After a quarter, it operates like a team member who has sat in every meeting and read every email. Persona memory builds over time. The more you use the Persona, the more context it retains about your preferences and working style. ## Your data, connected from day one A Persona is only as useful as the data it can reach. IdeaBoxAI connects directly to the systems your team already uses, so the Persona delivers answers grounded in live business data from the first conversation. | Connector | What the Persona can do with it | | ---------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Salesforce** | Pull live account data, pipeline status, opportunity details, contact records, and activity history. Surface deal risks and expansion signals without opening the CRM. | | **Google Drive** | Search and retrieve documents, templates, policies, and guides. Answer questions by citing specific sections from uploaded files. | | **Confluence** | Query internal documentation, knowledge articles, and onboarding content. Find answers buried in pages your team wrote but nobody remembers where. | | **Snowflake** | Translate natural language questions into SQL queries against your data warehouse. Return application pipelines, risk distributions, and trend analysis from structured tables. | | **Gong** | Access call summaries, transcript highlights, and conversation signals. Prepare for meetings using what was actually said in the last call, not what someone remembered to log. | | **Gmail** | Draft and send emails and meeting invites directly from the conversation. Every send requires explicit confirmation, so nothing goes out without your approval. | New connectors are added from **Settings > Connections**. Each integration is configured per-organization and scoped by persona permissions. Users only see data they are authorized to access. ## Not a chatbot. A team member. Generic AI tools answer questions. A Persona does work. It does not summarise what lien law says. It tells you which invoices have exposure this week and what to do about each one. It does not explain how to calculate deal health. It scores your pipeline, flags the three deals at risk, and drafts the follow-up email. It does not describe how AP triage works. It ranks your open invoices by urgency, checks for miscoded entries, and identifies the payments that will clear your lien exposure. The difference is specificity. Every answer is grounded in your data, shaped by your role, and designed for your workflow. Configure skills, connect data sources, assign users, and activate your Persona. Explore real deployments across construction, sales, insurance, and more. # Persona Setup Guide Source: https://docs.ideaboxai.com/copilot/persona-setup Understand the persona system, configure personas for different roles, attach skills and knowledge bases, and assign users to personas. A **Persona** defines how the Persona AI understands a user's role, goals, and the data sources they need. It shapes the intelligence surfaced, the starter prompts displayed, and the knowledge base scope. Users are assigned a default persona but can switch at any time. ## How personas work Each persona configures three key areas. * **Starter prompts**: Natural-language prompt cards shown on the home screen, based on the persona description and how the user interacts with the Persona AI. These help users get started quickly with common tasks for their role. * **Skills**: Purpose-built capabilities assigned to the persona that define what it can do. Skills handle specific tasks such as generating reports, drafting documents, or analyzing data. * **Knowledge base scope**: Which connected data sources and documents the persona can access and query. When a user switches personas, all three areas update immediately; no page refresh required. ## Supported personas The platform ships with built-in personas, and administrators can create custom personas for any role in the organization. Each persona is designed around a specific role, its goals, and the workflows it performs daily. A well-configured persona includes: * **Role definition**: A clear description of who the persona is for and what they do. * **Goals**: The outcomes the persona is optimized to support. * **Pain points addressed**: The manual or inefficient processes the persona eliminates. See the [Tutorials](/copilot/tutorials/sales-persona) and [Use Cases](/copilot/use-cases/business-outcome) sections for detailed examples of how specific personas are configured for different roles and industries. ## Navigate to persona management Open **Persona Management** from the left navigation. The Persona Management page lists every existing persona as a card. Use the **Search personas** box to filter by name and the **Newest first** dropdown to change the sort order. Each persona card shows the following. | Element | Description | | ------------------------ | ----------------------------------------------------------------------------- | | **Avatar and name** | The persona's coloured avatar initials and its name. | | **Active toggle** | Enable or disable the persona directly from the card. | | **Description** | A short summary of the persona's role and behaviour. | | **Assigned users** | Avatar chips for assigned users, with a **+ Add user** action to assign more. | | **Skills and knowledge** | Badges showing the number of skills and knowledge base items attached. | Persona Management page showing persona cards with avatars, Active toggles, descriptions, assigned users, and skill and knowledge base badges. ## Create or edit a persona Click **Create New Persona** in the top-right corner, or click the **Edit** action on an existing persona. When no personas exist yet, the page shows a **No personas yet** empty state with a **Create new persona** button. Persona Management empty state showing the No personas yet message and a Create new persona button. The Create New Persona modal contains the following fields. Fill in the following fields: | Field | Description | | ------------------------------------------- | --------------------------------------------------------------------------------------------------- | | **Name** | A short, descriptive persona name, for example, "Sales persona" or "Operations Manager". | | **How do you want this persona to behave?** | Describe the persona's focus areas and how it should respond. This drives skill selection and tone. | | **Assigned Users** | Search and select users who should have access to this persona. | Toggle **Enable Memory** to control whether the persona retains context across sessions. * **ON:** The Persona AI remembers previous conversations and builds on prior context. * **OFF:** Each session starts fresh with no retained history. Click **Create Persona** (or **Save Changes** when editing). The persona appears in the Persona Management list with an Active status by default. Assigned users will see it in their persona dropdown immediately. Create New Persona modal showing fields for Name, behaviour description, Assigned Users, Enable Memory toggle, and Features. ### Attach skills and knowledge bases Further down the Create New Persona modal, you can scope what the persona can do and what it can access. * **Skills**: **Select the AI capabilities this persona can use**, and search, and select skills. Selected skills appear on the right side, and the count is shown (for example, **ADDED (2)**). * **Knowledge bases**: **Attach knowledge bases this persona can search**; select the data sources the persona can query, and they appear on the right side. Create New Persona modal scrolled to the Skills section with a selected skill tag and the Knowledge bases section showing an attachable knowledge base. Once created, the personas appears as a card in the persona management page. ## Admin: assign users to a persona Personas are assigned to **existing platform users** directly from the Persona Management page. Each persona card includes a **+ Add user** action to grant additional users access to that persona. In the Admin Console, click **Personas** in the left sidebar. On the persona card you want to update, click **+ Add user**. Search for and select one or more existing users in the platform. Selected users appear as avatar chips on the persona card. Assigned users immediately gain access to the persona and see it in their persona dropdown; no page refresh required. Persona Management page showing a persona card with assigned user chips and the + Add user action used to assign existing platform users. # Settings and user preferences Source: https://docs.ideaboxai.com/copilot/settings Configure your profile and manage connections from the settings panel. Settings are accessible in the left-side bar. Navigate to Profile & Usage or Connections from here. ## Profile The Profile section manages your personal information and platform identity. The following settings are available. | Setting | Description | | ---------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | **Avatar** | 8 colour options. The selected colour is shown with a checkmark. The default is an initials-based avatar. | | **Display Name** | Editable. Used in greetings and references throughout the platform. | | **Personas** | Multi-select pill UI. Active personas are highlighted in blue. Drives starter prompts and knowledge-base scope. Available personas are configured by your administrator. | | **Language** | Dropdown. Default: English. Drives AI response language. | | **Last Active** | Read-only timestamp. | Profile settings panel showing avatar colour picker, display name field, persona pills, language dropdown, and last-active timestamp. ## Connections The Connections section manages all tool integrations, APIs, and database connections. A summary count of active connections is shown at the top. The following table describes the connection management interface. | Setting | Description | | ------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------- | | **Filter Tabs** | All \| MCP Integrations \| APIs \| Databases. Includes a search bar for name lookup. | | **Connected** | Card shows **Reconfigure** and **Disconnect** actions. Examples: Salesforce (API), Gong (API), Confluence (MCP), Data Warehouse (Database). | | **Available to connect** | Full-width **Connect** button. Examples: Google Drive (MCP), Gmail (API). | The following integration types are available. | Type | Integrations | | -------------------- | ------------------------------------------------------ | | **MCP Integrations** | Confluence, Google Drive, Slack, Notion, Jira, Linear. | | **APIs** | Salesforce, Gong, Gmail, HubSpot, Stripe, GitHub. | | **Databases** | Data Warehouse, PostgreSQL, MySQL. | Connections settings showing filter tabs, connected integrations with Reconfigure and Disconnect buttons, and available integrations with Connect buttons. # Share a persona Source: https://docs.ideaboxai.com/copilot/share-persona Share a persona with specific teammates or your whole organization by setting an access level and copying a share link. Once you've created a persona, you can share it with specific teammates or make it available to everyone in your organization. This guide covers setting the access level, adding individual people, and copying the share link. Only the persona's **owner** can change who it's shared with. ## Open the Share window Open the Share window from the persona's card on the Personas page. Click **Personas** in the left menu. Move your mouse over the card of the persona you want to share. A small toolbar appears at the bottom of the card. Click the **Share** icon (the arrow). The **Share \[Persona Name]** window opens. Persona card hover state showing the Share icon in the toolbar. ## Choose who can access it Under **Access**, click the dropdown to choose who can use this persona. | Access level | Who can use it | | ---------------- | ------------------------------------------------------- | | **Private** | Only you. No share link is created. | | **Organization** | Anyone in your workspace. | | **Public** | Anyone with the link, inside or outside your workspace. | Pick the level you want. The **Link** section at the bottom updates on its own. * **Private** - The link box shows "Private - no shareable link" and the Copy button is greyed out. * **Organization** or **Public** - A share link is created, and the **Copy** button becomes active. Access level dropdown showing Private, Organization, and Public options. ## Share with specific people If you'd like to share with certain teammates no matter which access level you chose, add them individually. Under **Share with people**, click the **Select teammates...** dropdown. A search box appears. Type a name or email to find someone, then choose one or more teammates from the list, for example, *Mohit Rohilla*, *Shreya Balaji*, and *Venkat Poleneni*. Click **Add**. The people you chose appear as small avatar icons under **Share with people**, and they can use the persona right away. Share with people dropdown showing teammate search and a list of teammates. ## Copy and share the link If you set the access to **Organization** or **Public**, copy the link and send it to your team. In the **Link** section, you see the share link, for example, `https://dev-refactor.ideaboxai.com/sha...`. Click **Copy** to copy the link. Share the link with your team through Slack, email, or wherever you like. Share dialog showing the shareable link and the active Copy button. When a persona is set to **Private**, the link box is greyed out and shows "Private - no shareable link". Change the access to **Organization** or **Public** to turn on link sharing. Share dialog in Private mode showing the disabled link field. ## What happens after you share After you save, sharing takes effect immediately. * The teammates you gave access to see the persona in their own **Personas** list. * A short message confirms your change was saved. * Teammates who don't have access to the underlying knowledge bases or connected tools may get more limited answers from the persona. Use **Organization** access when you're rolling a persona out to your whole team. Share it with a few key people first to check the answers before a wider rollout. # Share a skill Source: https://docs.ideaboxai.com/copilot/share-skill Share any skill with teammates or people outside your team using a link that lets anyone add the skill to their own workspace. You can share any of your skills with teammates or people outside your team using a link. Anyone who has the link can add the skill to their own workspace. This guide covers copying the share link, confirming it copied, and sending it on. ## Copy the share link Copy the link straight from the skill's card in Skills Studio. Go to **Skills Studio** by clicking **Skills** in the left menu. Move your mouse over the card for the skill you want to share. Click the **Share** icon (the arrow at the top of the card). A small label appears saying "Copy share link". Click it to copy the link. Skills Studio showing the Share icon tooltip on a skill card. ## Check that the link copied A message appears at the top of the screen to confirm the link is ready to paste. > **Share link copied.** Anyone with the link can add this skill. Toast notification confirming the share link has been copied. ## Send the link Paste the link into Slack, an email, or wherever you like. When someone opens it, they can add the skill straight to their workspace. Anyone with the link can add the skill, so only share it with people you want to give access to. Share skills across your teams so everyone works the same way and gets consistent results. To improve a skill after sharing, tell the Persona AI what you'd like changed and, it makes an updated version. # How to Create a Skill Source: https://docs.ideaboxai.com/copilot/skills-and-configuration Build and deploy AI skills that personas can use in conversations from the Skills Studio. Skills are the building blocks that make each persona capable. A skill defines a specific task the Persona AI can perform, such as generating a report, researching a topic, or drafting a document. Skills are managed from the **Skills Studio** in the Admin Console. ## Skills Studio overview Open the **Skill Studio** from the left sidebar. The dashboard displays all configured skills as a card grid. Each skill card shows: * **Skill name** and icon * **Description** of what the skill does * **Tags** indicating the skill type The top of the page shows the **Total Skills** and **Active** count. Use the **Enabled** and **Available** tabs to filter skills by type, and the search bar to find a specific skill by name. Skills Studio dashboard showing a grid of skill cards with names, descriptions, and tags, along with total and active skill counts. ## Create a skill There are two ways to create a skill: write instructions manually or upload a skill file. ### Write skill instructions Click **+ Create Skill** in the top-right corner, then select **Write skill instructions**. Complete the following fields: | Field | Description | | ---------------- | ----------------------------------------------------------------------------------- | | **Skill name** | A short, descriptive name for the skill. | | **Icon** | Choose an emoji to visually identify the skill. | | **Description** | Explain what the skill does and when it should be used. | | **Instructions** | Write the step-by-step instructions the AI should follow when executing this skill. | * **Grant trusted access**: Toggle on to allow the skill to execute actions directly without requiring user confirmation. Only enable this for skills you trust to run autonomously. * **Status**: Set to **Active** to make the skill available immediately, or leave it inactive until ready. Click **Save**. The skill appears in the Skills Studio grid and is available to the personas it is assigned to. Write skill instructions modal showing fields for Skill name, Icon, Description, Instructions, Grant trusted access toggle, and Active status. ### Upload a skill Click **+ Create Skill** in the top-right corner, then select **Upload a skill**. Drag and drop a skill file into the upload area, or click to browse your files. The following requirements apply: * **Supported formats:** JSON, YAML * **Maximum file size:** 3KB Click **Import** to add the skill to Skills Studio. The skill is parsed from the file and added to the grid. Upload skill modal with drag-and-drop area, file format requirements, and Import button. # About skills Source: https://docs.ideaboxai.com/copilot/skills-overview Learn what a skill is, how it connects to personas, and why reusable skills give you the same reliable result every time. A skill is a set of ready-made instructions you save once and reuse. It teaches a persona how to handle a specific task, like doing research, writing a report, or analyzing data, so you get the same reliable result every time, without having to explain the task again from scratch. For example, you could create a `market-competition-research` skill that always: * Looks up information about a market or competitor. * Reviews their positioning, pricing, strengths, and weaknesses. * Writes it all up in the same clear format: a summary, competitor profiles, and a side-by-side comparison. You build and manage skills in **Skills Studio**, then add them to one or more personas. ## How skills work with personas You add a skill to a persona from the **Edit Persona** panel. Once it's added, the persona knows how to do that task. It can start the task on its own when your message calls for it, or you can pick it yourself by typing `/` in the chat and choosing it from the list. The same skill can be added to as many personas as you like. For example, you could add one `market-competition-research` skill to both a Product Manager persona and a Sales persona. Both get the exact same research ability, while each keeps its own name, tone, and personality. Edit Persona panel showing the Skills section with Available and Added columns. To add a skill to a persona, follow these steps. Open the persona and click **Edit**. Scroll down to the **Skills** section. Look through the **Available** list, or search for the skill you want. Click the **+** next to a skill to move it into the **Added** column. Save the persona. The added skills are now available in that persona's chats. ## Why skills are useful Skills turn one-off instructions into a reusable capability your whole team can rely on. * **Same result every time** - Every persona that uses a skill follows the same steps and produces the same kind of output, no matter who's asking. * **Reuse instead of repeat** - Build a skill once and add it to as many personas as you need. You never have to write the same instructions twice. * **Easy for new people** - New team members can use proven, ready-made workflows from day one. * **Update in one place** - Change a skill once, and every persona that uses it is updated automatically. * **Works on its own** - A persona can tell when a skill fits what you're asking and use it for you, so you don't have to remember which skill to pick. Ready to build one? See [Create a skill with the Persona AI](/copilot/create-skill-ai) or [How to create a skill](/copilot/skills-and-configuration), then share it with your team using [Share a skill](/copilot/share-skill). # Using the Persona AI Source: https://docs.ideaboxai.com/copilot/tutorials/conversational-experience Learn how to interact with the personas, from sending prompts to generating documents and making calls from the Persona AI. The Persona AI is your primary interface for getting work done. This guide walks you through sending prompts, understanding response structures, using starter prompts, holding multi-turn conversations, generating documents, and making calls from the Persona AI. ## Sending your first prompt Navigate to the Persona AI home screen. You will see a chat bar at the bottom of the interface along with starter prompt cards above it. Click into the chat bar and type a natural-language question or instruction. For example: *"How is my pipeline this quarter?"* Click **Send** or press **Enter**. The AI processes your prompt and returns a structured response within seconds. Review the output: it will include data cards, tables, charts, or narrative text depending on your question. You do not need to use special syntax or commands. Just type what you need in plain language, and the AI will determine the best way to respond. ## Response structure Responses are not plain text. The Persona AI returns structured, data-driven output designed for fast consumption and action. * **Data cards**: Structured snapshots showing key metrics such as industry, revenue, employee count, deal stage, and account owner. * **Tables and charts**: Tabular data and visual charts for pipeline views, scoring breakdowns, and trend analysis. * **Source citations**: Every claim is grounded in connected data. Citations link back to the CRM record, document, or signal that informed the response. * **Suggested follow-ups**: The AI automatically suggests relevant next questions based on the response context, so you can drill deeper without starting over. ## Using starter prompts There are two ways to begin a task with the Persona AI. ### Click a starter prompt card On the home screen, click any of the pre-configured **starter prompt cards** displayed above the chat bar. Each card represents a common task for your persona. Clicking it pre-fills the prompt in the chat bar so you can review and send it. ### Type a prompt directly You can also type your request directly into the chat bar. The AI matches your intent to the appropriate skill and executes it. There is no need to memorize skill names or commands. Starter prompt cards are tailored to your active persona. If you do not see a relevant card, simply type your request directly in the chat bar. ## Making a call from the Persona AI The Persona AI can initiate calls directly from the conversation interface when connected to supported communication integrations. Type a prompt such as *"Call \[contact name]"* or *"Set up a call with \[contact name]."* The AI will look up the contact from your connected data sources. The AI displays the contact details and asks for confirmation before initiating the call. Review the details and confirm to proceed. Once confirmed, the call is initiated through the connected integration. The AI logs the call activity for your records. Call functionality requires a connected communication integration. Check your **Settings > Connections** to verify your integrations are configured. ## Multi-turn conversations The Persona AI retains context across turns within a conversation. You can ask follow-up questions, request refinements, or pivot to related topics without repeating earlier context. *"Pull the account brief for Acme Corp."* *"Now draft a follow-up email based on that briefing."* The AI uses the account brief from the previous turn to generate a personalized email. *"Add a reminder to follow up with their CTO next week."* The AI creates the task, still grounded in the same account context. Each conversation maintains its own context window. Starting a **new conversation** resets context. Use the same thread when you want the AI to remember what was discussed. ## Document generation The Persona AI can generate downloadable documents directly from a conversation. The following formats are supported. | Format | Use case | | --------------------- | --------------------------------------------------- | | **DOCX** | Proposals, account briefs, architecture summaries. | | **PDF** | Formatted reports, one-pagers, executive summaries. | | **Structured emails** | Drafts surfaced for review before sending. | Type a prompt such as *"Draft a proposal for the Acme Corp renewal"* or *"Generate a PDF summary of my pipeline."* The AI generates the document and displays a preview in chat. Review the content, structure, and data accuracy. Click **Download** to save the file locally, or click **Send** to deliver an email draft directly. You can edit the output before taking action. Documents are generated by combining templates from connected sources (such as Google Drive) with live data from your CRM and data warehouse. This means every document is current and personalized, not a static template. ## Example prompts by persona The following prompts show the types of questions different personas can ask. The Persona AI adapts depth, data sources, and output format based on the active persona. ### Account Executive * *"Pull the account brief for \[company]."* * *"How is my pipeline this quarter?"* * *"Draft a follow-up email for \[company]."* * *"Which deals are at risk?"* ### Sales Development Rep * *"Research \[company] before outreach."* * *"Draft outreach email for \[company]."* * *"Build a call script for \[company]."* * *"Which prospects should I prioritize?"* ### Sales Engineer * *"Generate discovery questions for \[industry]."* * *"Draft RFP response for \[requirement]."* * *"Configure demo for \[use case]."* * *"Handle \[technical objection]."* ### Admin * *"Show system usage this week."* * *"Create a new persona for \[role]."* # Sales Development Rep persona Source: https://docs.ideaboxai.com/copilot/tutorials/executive Set up and use the Sales Development Rep persona for prospect research, outreach generation, and pipeline building. The Sales Development Rep (SDR) persona is built for users who own top-of-funnel pipeline generation. It provides prospect enrichment, personalized outreach at scale, ICP scoring, and objection handling, enabling reps to go from 4-5 quality outreaches per day to 15-20 without sacrificing personalization. ## SDR use cases The following table summarizes the primary use cases for the SDR persona. | Use Case | Problem Solved | Key Output | | ------------------------------------- | -------------------------------------------------------- | ---------------------------------------------------------------- | | Prospect Research and Personalisation | Shallow research; outreach feels generic. | Auto-enriched profile: triggers, tech stack, pain points, hooks. | | Outreach Email Drafting | Stale templates; low reply rates. | AI-personalised emails with A/B variants. | | ICP Scoring and Prioritisation | All leads treated equally; reps chase low-fit accounts. | AI-scored ICP fit with ranked lists. | | Call Script Generator | No standardised call framework; new reps wing it. | Dynamic scripts adapted to prospect role, industry, and stage. | | Objection Handling Guide | Reps freeze on objections; responses vary by experience. | Real-time objection library with rebuttals. | | Follow-up Sequence Creator | Follow-up is ad hoc; ghost rate high. | Multi-step sequences timed to engagement signals. | ## Set up the SDR persona In the Admin Console, click **Persona Management** and then **+ Create New Persona**. * **Name:** SDR * **Behavior:** "You are a Sales Development Rep. Help research prospects, generate personalized outreach, score ICP fit, build call scripts, handle objections, and create follow-up sequences. Ground all responses in CRM data, data warehouse signals, and public web intelligence." * **Enable Memory:** ON Enable **Access Connections**, **Access Knowledge Base**, **Use Email Tools**, and **Browse Web** (SDRs need web research for prospect enrichment). Assign the following skills to the persona: * Prospect Research * Email Drafting * ICP Scoring * Call Script Generator * Objection Handling Add starter prompts tailored to the SDR role: * "Research \[company name] before outreach." * "Draft outreach email for \[company name]." * "Build a call script for \[company name]." * "Handle 'we already have a tool' objection." * "Which prospects should I prioritize this week?" * "Build follow-up sequence for \[contact name] at \[company name]." Assign your SDR team members to this persona. They will see the SDR-specific home screen on their next login. ## Walkthrough: prospect research This walkthrough demonstrates how the SDR persona handles a prospect research request end to end. Open the Conversational AI and type: *"Research \[company name] and give me cold outreach angles for their Head of Data Platform."* The AI returns a multi-section prospect profile: * **ICP Score**: A fit rating with rationale based on CRM data, hiring signals, and tech stack analysis. * **Company snapshot**: Industry, revenue, employee count, recent funding, and key initiatives. * **Trigger events**: Leadership changes, product launches, earnings calls, job postings, and expansion signals relevant to your value proposition. * **Pain points and hooks**: Identified challenges mapped to your solution's strengths, ready to use in outreach. Persona AI researching a prospect and returning a structured snapshot with positioning, use cases, and team signals. Click a suggested follow-up such as *"Draft the cold email"* or *"Write a cold call script"*, or type your own follow-up prompt. The AI retains the full prospect context. ## Walkthrough: outreach drafting This walkthrough shows how to generate personalised outreach emails from a prospect research result. After completing prospect research (or in a new conversation), type: *"Draft outreach email for \[company name]."* The AI produces a personalized email that includes: * **Subject line**: Tailored to the prospect's trigger events or pain points. * **Opening hook**: A personalized reference grounded in real signals (not generic flattery). * **Value proposition**: Mapped to the prospect's specific challenges and tech stack. * **Call to action**: A clear, low-friction next step. * **A/B variant**: An alternative version with a different angle for testing. Review the draft, make any adjustments, and click **Send** to deliver the email or **Copy** to use it in your email client. When you generate outreach from within the same conversation as a prospect research query, the AI uses the full enriched profile to personalize the email. This produces significantly better results than drafting in a fresh conversation. ## Example starter prompts The following starter prompts are recommended for the SDR persona. Each one maps to a common SDR task. * *"Research \[company name] before outreach."* * *"Draft outreach email for \[company name]."* * *"Build a call script for \[company name]."* * *"Handle 'we already have a tool' objection."* * *"Which prospects should I prioritize this week?"* * *"Build follow-up sequence for \[contact name] at \[company name]."* # Sales Engineer persona Source: https://docs.ideaboxai.com/copilot/tutorials/operations-leader Learn how to configure the Sales Engineer persona for technical discovery, demo prep, RFP responses, and POC planning. In this tutorial, you will configure the Sales Engineer (SE) persona and walk through three common workflows: demo preparation, RFP response drafting, and prospect research. The SE persona is designed for users who own technical credibility and POC success. It provides structured discovery questions, demo configuration, RFP response drafting, and POC planning. ## What the SE persona provides The following table summarizes the primary use cases for the SE persona. | Use Case | Problem Solved | Key Output | | ---------------------------------- | ------------------------------------------------------------------------- | ------------------------------------------------------------------------- | | Technical Discovery Questions | Discovery depth varies by SE experience; new SEs miss critical questions. | AI-generated discovery sets by industry, tech stack, and deal stage. | | RFP / RFI Response Assistant | RFP responses take days; rewritten from scratch each time. | AI-drafted responses from approved answer library; gap flagging. | | Demo Environment and Script Config | Manual demo prep; generic demos don't resonate. | Auto-configured demo scripts with environment setup checklists. | | Technical Objection Handling | Technical objections stall deals; responses vary by SE. | Context-aware objection library with evidence-based responses. | | POC Planning and Tracking | POC scope creep; success criteria unclear; no standard tracking. | Structured POC plans with competitive-aware scoping and milestone alerts. | | Solution Architecture Summaries | Arch docs created ad hoc; knowledge lost when SEs move between deals. | Auto-generated architecture summaries: stack, integrations, data flows. | ## Set up the SE persona In the Admin Console, click **Persona Management** and then **+ Create New Persona**. * **Name:** Sales Engineer * **Behavior:** "You are a Sales Engineer. Help run structured technical discovery, prepare tailored demos, respond to RFPs and technical objections, plan and track POCs, and generate solution architecture documents. Ground all responses in internal knowledge bases, CRM data, and prospect context." * **Enable Memory:** ON Enable **Access Connections**, **Access Knowledge Base**, and **Browse Web** (SEs need web access for prospect tech stack research). Disable **Use Email Tools** unless SEs send directly from the platform. Assign the following skills: Technical Discovery, RFP Response, Demo Prep, Technical Objection Handling, POC Planning. Add starter prompts tailored to the SE role: * "Generate discovery questions for \[industry] prospect." * "Prep me for tomorrow's demo with \[company name]." * "Draft an RFP response for \[topic]." * "What technical objections should I expect?" * "Build a POC plan for \[company name]." * "Create a solution architecture summary for \[deal name]." Assign your SE team members to this persona. ## Example: demo preparation This example shows how the SE persona generates a tailored demo brief from a single prompt. *"Prep me for tomorrow's technical demo with Apex Dynamics. They are evaluating us against Snowflake for their data platform consolidation."* The Persona AI returns the following sections. * **Technical context**: Current stack, evaluation driver, employee count, and deployment requirements. * **Likely technical objections**: Specific objections mapped to the prospect's stack, with recommended responses and proof points. * **Recommended demo flow**: A structured demo sequence tailored to the prospect's evaluation criteria and competitive position. Persona AI generating a full demo prep plan with outcomes, a timed run of show, and objection handling. ## Example: RFP response This example shows how the SE persona drafts an RFP response from your approved answer library. *"Draft an RFP response for data integration capabilities"* The Persona AI queries the approved answer library in your connected knowledge base and drafts a response following your standard format. If the answer library does not cover a question, then the Persona AI flags the gap. The SE can review, edit, and export the response as a DOCX document. ## Example: prospect research This example shows how the SE persona researches a prospect and generates cold outreach angles grounded in public data. *"Research ideaboxAI and give me cold outreach angles for their Head of Data Platform"* The Persona AI runs a web search and crawls the prospect's public pages to gather product positioning, tech stack signals, and infrastructure context. It then generates targeted outreach angles, each with a hook, talk track, and call to action. Persona AI generating cold outreach angles grounded in the prospect's public data and positioning. ## Next steps Now that the SE persona is configured, explore related tutorials to expand your setup. Learn how to send prompts, hold multi-turn conversations, and generate documents. Configure the AE persona for deal intelligence, meeting prep, and pipeline management. Set up the SDR persona for prospect research and outreach generation. Learn how to set up and manage data source connectors for your workspace. # Account Executive persona Source: https://docs.ideaboxai.com/copilot/tutorials/sales-persona Learn how to configure the Account Executive persona for deal intelligence, meeting prep, and pipeline management. In this tutorial, you will configure the Account Executive (AE) persona, connect it to Google Drive, and walk through three common workflows: accessing sales folders, preparing for meetings, and sending follow-up emails. The AE persona is built for users who own pipeline and revenue. It surfaces deal intelligence, account briefings, competitive context, and relationship insights, reducing research time from 30 to 60 minutes to under 5 minutes per account. ## AE use cases The following table summarizes the primary use cases for the AE persona. | Use Case | Problem Solved | Key Output | | ----------------------------- | ----------------------------------------------------------------- | -------------------------------------------------------------------------- | | Account Research and Briefing | 30โ€“60 min manual research before every engagement. | Auto-generated value brief: snapshot, priorities, tech stack, talk tracks. | | Deal Coaching and Next Steps | Deal risk invisible; missing artefacts undetected until too late. | Deal health scorecard, risk flags, next best action, momentum alerts. | | Meeting Prep | Generic prep regardless of audience; AE/SE misaligned. | Audience-aware brief per attendee role; unified prep sheet. | | Proposal and Email Drafting | Proposals take hours; follow-ups delayed; copy-paste errors. | Template auto-population from CRM; post-meeting email drafts. | | Competitive Intelligence | Battlecards static and not deal-specific; no pre-call framing. | Deal-specific competitive briefs; pre-call positioning notes. | | Exec Stakeholder Mapping | Single-threaded deals; no visibility into buying committee gaps. | Dynamic stakeholder map, buying committee gap analysis. | ## Set up the AE persona In the Admin Console, click **Persona Management** and then **+ Create New Persona**. * **Name:** Account Executive * **Behavior:** "You are an Account Executive. Help manage accounts, track deal health, prepare for meetings, generate proposals, and surface competitive intelligence. Ground all responses in CRM data, data warehouse signals, and internal knowledge." * **Enable Memory:** ON Enable **Access Connections**, **Access Knowledge Base**, and **Use Email Tools**. Disable **Browse Web** unless needed. Assign the following skills to the persona: * Account Research * Deal Analysis * Email Drafting * Meeting Prep * Competitive Intelligence Add starter prompts tailored to the AE role: * "Show me my current pipeline." * "Which deals are at risk of slipping?" * "Prep me for my QBR with \[account name]." * "Draft a follow-up email for \[account name]." * "How do we compare to competitors?" * "Am I on track to hit my quota?" * "Show me my top accounts by revenue." Assign your AE team members to this persona. They will see the AE-specific home screen on their next login. ## Walkthrough: accessing your sales folders This walkthrough demonstrates how the AE persona connects to Google Drive to access your account folders. Before the Persona AI can access your files, Google Drive must be connected to the current session. If you ask a Drive-related question without a connection, then the Persona AI prompts you to connect. Persona AI prompting to connect Google Drive from the Connections panel. Open the **Connectors** panel from the bottom-left of the Persona AI and toggle **Google Drive** on. The connector must be enabled for the session before the Persona AI can search, list, or read files from your Drive. Connectors panel with Google Drive toggled on for the current session. If the folder you need is in a **Shared drive**, then let the Persona AI know so it can locate it correctly. To learn more about setting up and managing connectors, see [Connectors overview](/connectors/overview). Once Google Drive is connected, type: *"List the folders inside my Google Drive folder called 'Sales Persona AI (AE-01)'"* The Persona AI searches your Drive, locates the folder, and returns a list of all subfolders with direct links. Persona AI listing all account subfolders inside the Sales Persona AI folder with direct links. Click any subfolder name or link to open it directly in Google Drive. You can also ask the Persona AI to list the files inside a specific account folder. ## Walkthrough: meeting prep This walkthrough shows how the AE persona prepares you for an upcoming meeting. The Persona AI pulls from your Drive files, runs web research, and generates a tailored briefing. Type a prompt that includes the account name, meeting context, and any relevant details: *"Prep me for my first meeting with Meridian Health Systems next Tuesday. Healthcare company, \$200K opportunity, Stage 1. I've never engaged with them before."* The Persona AI automatically: * **Searches Google Drive** for relevant files in your account folders, such as case studies and benchmark decks. * **Retrieves file content** from matched documents to extract key talking points. * **Runs a web search** on the account to surface recent news, company profile, industry context, and key contacts. Persona AI pulling case studies from Google Drive and running a web search to build a meeting brief. The response includes: * **Company overview**: Industry, key roles, and business context. * **Internal assets**: Relevant case studies, benchmark decks, and collateral from your Drive. * **Suggested talk track**: An opening approach and top questions tailored to the account's industry and deal stage. Use the suggested talk track directly or ask follow-up questions such as *"What objections should I expect?"* or *"Draft the follow-up email."* The Persona AI retains the full meeting context. ## Walkthrough: email follow-up This walkthrough shows how to draft and send a follow-up email directly from the Persona AI after a meeting prep or account interaction. After completing a meeting prep or account research, type: *"Draft the follow-up email to \[contact name]. Send the email to \[email address]."* The Persona AI generates a follow-up email and opens the **Email Composer**. The draft includes: * **To**: Pre-filled with the contact's email address. * **Subject line**: A clear subject referencing the account and proposed next step. * **Body**: A professional follow-up grounded in the meeting context, including action items and next steps. Email Composer showing a draft follow-up email ready for review before sending. Review the draft in the Email Composer. Click **Edit draft** to make changes, **Cancel** to discard, or **Send Email** to deliver it directly from the Persona AI. Drafting an email in the same conversation as a meeting prep lets the Persona AI use the full account briefing to personalize the message. This produces better results than drafting in a fresh conversation. ## Next steps Now that the AE persona is configured, explore related tutorials to expand your setup. Learn how to send prompts, hold multi-turn conversations, and generate documents. Configure the SE persona for technical discovery, demo prep, and POC planning. Set up the SDR persona for prospect research and outreach generation. Learn how to set up and manage data source connectors for your workspace. # Sales Team Source: https://docs.ideaboxai.com/copilot/use-cases/actian-sales-copilot-case-study Learn how IdeaBoxAI's Sales Team Persona AI accelerates deal velocity across account research, deal health, technical discovery, and outbound prospecting, powered by 32 purpose-built skills. Every Account Executive knows the Sunday evening dread. Eight deals on the forecast. Two have gone quiet. One has a close date this Friday that everyone knows is slipping. Your manager wants a pipeline update at 9am Monday, and you still have not updated Salesforce from last week's calls. Every Sales Engineer knows the 4 pm scramble. A discovery call just ended. The prospect shared data volumes, query times, and stack details. You need to capture all of it, assess fit, and scope a POC before you forget what was said. Instead, you open a blank document and start typing from memory. Every SDR knows the Monday morning paralysis. Forty-seven leads in the queue. No signal on which ones are worth calling first. You pick one at random, spend 20 minutes researching, draft an email, and hope it lands. Multiply that by five and your morning is gone. The Sales Team Persona AI solves all three. It connects directly to Salesforce, Confluence, Google Drive, LinkedIn, and the public web. It understands your role, pulls live data, and answers your actual questions, in seconds, not hours. ## The Persona AI knows your role before you ask A Persona in IdeaBoxAI is the AI's understanding of who you are, what you need, and what your data looks like. When you log in as an Account Executive, you get a deal-management intelligence layer tuned to pipeline health, meeting prep, and forecast accuracy. When you log in as a Sales Engineer, you get a technical layer tuned to discovery, POC design, and objection handling. The Sales Team Persona AI ships with three purpose-built personas. Each connects to the data sources relevant to their role. The persona you are assigned to determines which data you see, which skills are available, and which scenario cards appear on your home screen. The following table summarizes the three personas. | **Persona** | **Focus** | **Key Needs** | | --------------------------- | ---------------------------- | ---------------------------------------------------------------------------------------------------- | | Account Executive (AE) | Deals, meetings, pipeline | Account research, deal health, meeting coaching, forecast prep, Salesforce hygiene | | Sales Engineer (SE) | Discovery, POC, objections | Discovery questions, technical fit assessment, POC scoping, competitive rebuttals, evidence assembly | | Sales Development Rep (SDR) | Leads, outbound, prospecting | ICP scoring, prospect profiling, outreach sequences, lead segmentation, Salesforce sync | Each persona also comes with pre-configured scenario cards on the home screen. Click any card to pre-fill the chat input with a common workflow prompt. You can also type your request directly in plain language; the Persona AI auto-detects the right skill. To browse all available skills, type `/` in the chat bar to open the skill picker. ## AE scenario: account research before a first meeting **Jordan Reeves ยท Account Executive** Mid-market territory ยท 14 active opportunities ยท First call with Meridian Health tomorrow at 10 am Jordan has a first meeting with Meridian Health in the morning. Normally, this means 45 minutes of tab-switching, LinkedIn for the org chart, the company website for strategic priorities, Salesforce for prior activity, and Google for recent news. By the time the research is done, there is no time left to build a talk track. **Jordan types:** *"Prep me for my meeting with Meridian Health tomorrow."* **The Persona AI responds:** > **Account brief, Meridian Health** > Prepared for Jordan Reeves ยท Meeting: Apr 24, 10:00 am ยท First call > > **Company snapshot** > > | **Attribute** | **Detail** | > | ------------------------- | ------------------------------------------------------------------- | > | Industry | Healthcare / Hospital systems | > | HQ | Minneapolis, MN | > | Revenue | \$2.1B (FY25) | > | Employees | 8,400 | > | Data infrastructure spend | Estimated \$14M annually (based on job postings and vendor signals) | > > **Strategic priorities and signals** > > * Announced a \$40M data modernization initiative; in Q1 earnings call, migrating from on-prem Teradata to a hybrid cloud architecture. > * Posted 6 data engineering roles in the last 90 days. Three mention "real-time analytics" and two mention "query performance at scale." > * New VP of Data Platform (Anika Patel) started 4 months ago. LinkedIn shows she came from UnitedHealth, where she led a similar migration. > > **Tech stack gaps** > > | **Current stack** | **Signal source** | **Actian fit** | > | ---------------------- | ------------------------------ | ----------------------------------------------------------------------------------- | > | Teradata (on-prem) | Job postings, Confluence notes | ๐ŸŸข High, migration path from Teradata is a proven use case. | > | Snowflake (evaluation) | Salesforce competitor field | ๐ŸŸก Moderate, they are evaluating, not committed. Real-time query gap is the angle. | > | Informatica ETL | Job postings | ๐ŸŸข High, DataConnect replacement opportunity if pipeline modernization is in scope. | > > **Competitive context** > Snowflake is listed as a competitor on this opportunity in Salesforce. Based on their data volumes (likely 500M+ rows given their patient record count) and their real-time requirements, the positioning angle is query latency at high concurrency, not migration ease. > > **Recommended talk track** > Lead with the data modernisation initiative, it is their stated priority, not something you are inferring. Acknowledge that Snowflake is strong for warehousing, then pivot to the real-time gap: "The question is not whether Snowflake is a good warehouse; it is whether it is the right engine for sub-second queries at your projected data volumes." > > Reference the FinServ customer running 3.8PB with sub-500ms average query time; healthcare and financial services share the same concurrency and compliance requirements. > > **Suggested questions for the call** > > * "What does your timeline look like for the Teradata migration, and what is driving the deadline?" > * "When you say real-time analytics, what query response time does your clinical team need to make decisions?" > * "How are you handling the compliance side of the migration? Is HIPAA driving any of your architecture choices?" > > **Suggested follow-ups:** Align my SE for this meeting ยท Build a pre-meeting checklist ยท Show me similar closed-won deals in healthcare ยท Save this brief to Salesforce Jordan reviews the brief in 4 minutes. The talk track is ready. The competitive angle is specific. The questions are tailored to what the prospect actually cares about, not a generic discovery script. **Jordan types:** *"I'm bringing my SE to this meeting. Brief my SE."* **The Persona AI responds:** > **SE alignment brief, Meridian Health** > For: SE assigned to Meridian Health ยท Meeting: Apr 24, 10:00 am ยท Stage 1 (First Call) > > **Account context** > > * Healthcare system, \$2.1B revenue, migrating from Teradata to hybrid cloud. > * Evaluating Snowflake. New VP of Data Platform (Anika Patel) is driving the initiative. > * Key signal: 6 data engineering hires in 90 days, 3 mention real-time analytics. > > **Your role in this meeting** > > * Listen for data volume numbers, current query times, and concurrency requirements. These are the inputs for the competitive angle. > * Be ready to address the Snowflake comparison, but do not lead with it. Wait for them to bring it up or for Jordan to set up the positioning. > > **Capabilities to demonstrate** > > * Vector AI DB real-time query performance at high row counts, the Teradata replacement story. > * DataConnect as an Informatica ETL alternative, only if they raise pipeline modernization. > > **POC design outline** > If the call goes well and they share specific data volumes, a 2-week POC on their patient analytics dataset would demonstrate the latency gap. Target metric: sub-2s query on a 30-day rolling window across their projected row count. > > **What NOT to lead with** > Do not lead with pricing or benchmarks in a Stage 1 call. Do not position against Snowflake unless they raise it first. Do not suggest a POC until discovery is complete. The SE alignment brief reuses research from the account brief generated in the same session. The Persona AI does not repeat the full data pull; it builds on what it already knows. The results: * Account research reduced from 45 minutes to under 5 minutes. * Talk track tailored to the prospect's stated priorities, competitive situation, and tech stack, not a generic pitch. * SE briefed with role-specific context and a clear POC outline in 30 seconds. * All research auto-saved to Salesforce and Google Drive. ## AE scenario: deal health check before forecast **Rachel Torres ยท Account Executive** Enterprise territory ยท 11 active opportunities ยท Forecast call with VP at 2 pm Rachel has a forecast call in two hours. She knows her pipeline has problems but cannot pinpoint which deals are actually at risk versus which ones just feel slow. Before IdeaBoxAI, she would open each opportunity in Salesforce one by one and try to piece together the story from activity logs and her own memory. **Rachel types:** *"Run a deal health check on NovaTech. They've gone quiet, and I'm worried about the close date."* **The Persona AI responds:** > **Deal health scorecard, NovaTech** > Opportunity: NovaTech, Enterprise Data Platform ยท ARR: \$185,000 ยท Stage 4 ยท Close date: May 16 > > **Overall score: 54 / 100, Recovering** ๐ŸŸก > > | **Dimension** | **Score** | **Evidence** | > | --------------------- | ----------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | > | Activity momentum | 35 / 100 ๐Ÿ”ด | Last activity was 18 days ago (email from Rachel, no reply). Prior to that, 3 calls in 2 weeks. The drop is sharp. | > | Stakeholder coverage | 45 / 100 ๐Ÿ”ด | Only 1 contact engaged (David Cho, Sr. Data Engineer). No executive sponsor identified. No procurement contact. Single-threaded. | > | Artifact completeness | 70 / 100 ๐ŸŸก | Account brief on file. Discovery notes on file. Missing: mutual action plan, proposal, executive sponsor alignment. | > | Close date integrity | 50 / 100 ๐ŸŸก | Close date has been pushed once already (from Apr 30 to May 16). At current velocity, May 16 requires a stage progression within 5 business days, which has not happened. | > | Competitive exposure | 65 / 100 ๐ŸŸก | Databricks listed as competitor. No update on competitive status since Stage 2. Last mention: "They're also talking to Databricks but haven't started a POC." That was 4 weeks ago. | > > **Risk flags** > > * **Stale activity (critical):** 18 days since last response. Your average deal at Stage 4 has contact every 4.2 days. NovaTech is 4x the norm. > * **Single-threaded (high):** Only David Cho is engaged. He is a Sr. Data Engineer, not a decision-maker. No executive sponsor on record. > * **Close date drift (medium):** Already pushed once. Current trajectory does not support May 16 without immediate re-engagement. > > **Benchmark:** Deals at the same stage and ARR that closed-won had an average health score of 78. NovaTech is 24 points below that benchmark. > > **Suggested follow-ups:** View next actions for this deal ยท Draft a re-engagement email for David Cho ยท Generate a mutual action plan ยท Update Salesforce with forecast category Rachel clicks the suggested follow-up. **Rachel types:** *"What should I do today to save this deal?"* **The Persona AI responds:** > **Next actions, NovaTech** > Based on deal health score (54/100) and 3 open risk flags. > > **Today**
# Action Channel Why Exact words
1 Re-engage David Cho Call (not email) 18 days silent. Email has failed; David has not replied to the last 2. A call breaks the pattern. "David, I wanted to check in directly. Last time we spoke, you were comparing timelines with your Databricks evaluation. I have some new benchmark data that might help with that. Do you have 10 minutes this week?"
2 Ask David for an exec intro Same call Single-threaded risk. You need a second contact before this deal can progress. "One thing that would help me build the best proposal for your team: would it make sense to include your VP in a 15-minute alignment call? I want to make sure the business case lands, not just the technical one."
3 Update Salesforce Persona AI Log today's call attempt, update next step date, flag the single-threaded risk in the opportunity notes. I can update Salesforce for you after the call; just tell me what happened.
> **This week**
# Action Goal
4 Generate a mutual action plan working back from May 16. Creates a shared timeline that forces both sides to commit to milestones, or reveals that May 16 is not realistic.
5 Draft a re-engagement email with the Databricks benchmark as the hook. Backup if the call does not connect. Subject references the benchmark, not "checking in."
6 Update forecast category to Best Case with a note. Score is 54 with active re-engagement. Commit requires a score above 90 with no open flags.
Rachel walks into her forecast call with a clear story: NovaTech is at risk, here is why, here is what she is doing about it today, and here is what the forecast category should be, with evidence behind each point. The results: * Deal risk surfaced across 5 dimensions with specific Salesforce evidence, not gut feel. * 18-day silence flagged and benchmarked against similar deals at the same stage. * Prioritized action plan split into today and this week, with exact words for each conversation. * Forecast category recommended with rationale the VP can audit. ## SE scenario: post-discovery debrief and POC scoping **Marcus Liu ยท Sales Engineer** Assigned to 6 active evaluations ยท Just finished a discovery call with DataStream Analytics Marcus just got off a 45-minute technical discovery call with DataStream Analytics. The champion shared volumes, query requirements, stack details, and a decision timeline. Marcus needs to capture all of it before the details blur, and assess whether this is a strong fit. **Marcus types:** *"Call done, here's what I learned. DataStream Analytics, VP of Engineering is the champion. They're running Redshift today with about 800M rows, growing 15% quarterly. Current dashboard queries take 12-18 seconds; they need sub-2 seconds. They're evaluating us alongside BigQuery. Decision by end of June, POC needs to start by mid-May. Stack is Python, Airflow, dbt. Main concern is migration complexity from Redshift."* **The Persona AI responds:** > **Post-discovery summary, DataStream Analytics** > > **Key facts card** > > | **Attribute** | **Detail** | > | ------------------- | --------------------------------- | > | Company | DataStream Analytics | > | Champion | VP of Engineering | > | Current stack | Redshift, Python, Airflow, dbt | > | Data volume | 800M rows, growing 15% quarterly | > | Current query time | 12โ€“18 seconds (dashboard queries) | > | Required query time | Sub-2 seconds | > | Competitor | BigQuery | > | Decision deadline | End of June | > | POC start window | Mid-May | > > **Actian fit assessment: Strong** ๐ŸŸข > > The fit is strong for three reasons: > > * Their query performance gap is significant. 12โ€“18 seconds down to sub-2 seconds is exactly the use case where Vector AI DB outperforms both Redshift and BigQuery at their row count. > * 800M rows growing 15% quarterly puts them at 1.2B rows within 12 months. BigQuery charges per query at that scale. Vector AI DB has no per-query cost; the TCO argument strengthens as they grow. > * Their stack (Python, Airflow, dbt) integrates cleanly. No exotic dependencies. > > **Migration risk note:** The champion flagged migration complexity as their main concern. This is a category 2 objection (Migration Risk). Acknowledge the real cost first; do not minimize it. Then reframe: the question is not whether migration has a cost, but whether staying on Redshift at 12โ€“18 second query times has a higher one. > > **Recommended next step:** Generate a POC proposal scoped to their dashboard use case. Target metric: sub-2s query on a 30-day rolling window at 800M rows. > > **Suggested follow-ups:** Generate a POC proposal for DataStream ยท Draft an email to the champion with POC scope ยท Update Salesforce stage ยท Pull benchmark data for Redshift vs Vector AI DB Marcus clicks the POC suggestion. **Marcus types:** *"Generate a POC proposal for DataStream."* **The Persona AI responds:** > **POC scope card, DataStream Analytics** > > | **Attribute** | **Detail** | > | --------------------- | --------------------------------------------------------------------------------------------- | > | Dataset | Production dashboard dataset, live Redshift export (preferred over synthetic). | > | Dataset size | 800M rows current, test at 1.2B to demonstrate growth headroom. | > | Success metric | Sub-2s query on 30-day rolling window aggregation at 800M rows. Stretch: sub-1s at 1.2B rows. | > | Champion | VP of Engineering | > | Duration | 2 weeks (standard, integration complexity is low given Python/Airflow/dbt stack). | > | DataConnect | Not included. Their Airflow pipeline handles ingestion today; no ETL replacement needed. | > | Vector AI DB use case | Real-time dashboard analytics replacing Redshift. | > > **Suggested next steps** > > 1. Send the POC scope to the champion for sign-off on dataset and success metric. > 2. Schedule a 30-minute POC kickoff call with Marcus and the champion's engineering lead. > 3. Request a Redshift export of 30 days of dashboard query data for baseline comparison. > 4. Draft a post-call email to the champion confirming the POC timeline and what both sides need to deliver. The POC proposal pulls directly from the discovery debrief; dataset size, query requirements, and stack details are carried forward automatically. The success metric is quantified, not qualitative. The results: * Discovery findings captured and structured in under 60 seconds, no manual note formatting. * Actian fit scored with specific reasoning tied to data points from the call, not generic categories. * POC scoped with a quantified success metric, realistic timeline, and clear next steps. * All findings auto-logged to Salesforce. ## AE scenario: post-meeting log and follow-up email **Sarah Park ยท Account Executive** Enterprise territory ยท Just finished a Stage 3 call with Vertex Financial Sarah just walked out of a meeting with Vertex Financial. She has 12 minutes before her next call and needs to log what happened, update Salesforce, and send a follow-up email. Before IdeaBoxAI, two of those three would not get done until the end of the day, if at all. **Sarah types:** *"Meeting done with Vertex Financial, here's what happened. Good energy overall. CFO joined for the first 15 minutes, which was unexpected. She asked about SOC 2 compliance and whether we support data residency in the EU. I confirmed SOC 2 Type II and said I'd follow up on EU residency. David (champion) confirmed they want to move to POC but need budget approval from the CFO first. Competitor is Snowflake; they've done a POC with Snowflake already but weren't happy with query latency on their risk modeling workload. Next step is a technical deep-dive with their data engineering team next week. One risk: David mentioned the project could get deprioritized if Q2 revenue comes in below target."* **The Persona AI responds:** > **Post-meeting log, Vertex Financial** > Meeting date: Apr 23, 2026 ยท Stage 3 ยท Attendees: Sarah Park (AE), David (Champion), CFO (new contact) > > **What went well** > > * CFO joined unprompted, signals executive interest. This is unusual at Stage 3 and suggests the initiative has visibility at the C-level. > * Champion confirmed intent to move to POC. Budget approval is the remaining gate. > * Snowflake POC completed but did not meet latency requirements on risk modelling. This is a significant competitive advantage; they have already experienced the gap you are positioning against. > > **Deal status** > Stage stays at 3. Move to Stage 4 when budget approval is confirmed, and the technical deep-dive is scheduled with a date. > > **Open risks** > > | **Risk** | **Severity** | **Specific action** | **Exact words** | > | ------------------------------------- | ------------ | ----------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------- | > | EU data residency question unanswered | ๐ŸŸก Medium | Follow up within 24 hours. Delayed response on a CFO question erodes trust. | Include in the follow-up email with a clear answer or a specific date when you will have one. | > | Budget deprioritization if Q2 misses | ๐ŸŸก Medium | Accelerate the timeline. The longer this deal sits at Stage 3, the higher the risk of a freeze. | In your next call with David: "What does the budget approval process look like, and is there a way to get sign-off before the Q2 review?" | > | CFO is a new stakeholder, no record | ๐ŸŸข Low | Add the CFO as a contact in Salesforce today. Log her SOC 2 and residency questions. | I can add her to the opportunity record when you confirm the update. | > > **Immediate next actions** > > 1. Send follow-up email to David confirming the technical deep-dive and addressing the EU residency question. (Due: today.) > 2. Add CFO as a contact in Salesforce with meeting notes. (Due: today.) > 3. Schedule the technical deep-dive with David's data engineering team. (Due: this week.) > > **Suggested follow-ups:** Draft the follow-up email ยท Update Salesforce with these notes ยท Generate a mutual action plan for Vertex ยท Prepare objection prep for the next call Sarah immediately follows up. **Sarah types:** *"Draft the follow-up email to David."* **The Persona AI responds:** > **Email draft, Post-meeting follow-up** > > **To:** David ยท Vertex Financial > **Subject:** Vertex risk modeling POC, next steps and residency answer > > David, > > Good speaking with you and your CFO today. Two follow-ups from the conversation. > > First, on EU data residency, I am confirming the specifics with our compliance team and will have a documented answer for you by Friday. SOC 2 Type II certification is current, and I can share the report directly if helpful for your CFO's review. > > Second, the technical deep-dive. Are you available next Tuesday or Wednesday for a 60-minute session with your data engineering team? Marcus (our SE) will walk through the architecture and how it maps to your risk modeling workload, specifically the latency gap you experienced in the Snowflake POC. > > Let me know what works. > > Sarah The Persona AI requires explicit confirmation before writing anything to Salesforce. It shows you exactly what will be logged: activities, notes, stage changes, and forecast updates, and waits for your approval. The results: * Meeting logged with structured outcomes, risk flags, and next actions in under 90 seconds. * Follow-up email drafted with a specific hook (the Snowflake latency gap) and a concrete ask (deep-dive date), no generic opener. * CFO flagged as a new stakeholder with her specific concerns captured for future reference. * 12-minute window between meetings used productively instead of lost. ## SDR scenario: weekly lead scoring and outbound **Priya Mehta ยท Sales Development Rep** Assigned territory: West Coast mid-market ยท 47 unworked leads in Salesforce Priya starts every Monday the same way, staring at a list of leads with no idea which ones to call first. She used to pick accounts alphabetically or by the order they came in. Half her outreach went to companies that were never going to buy. The other half went out with a generic template. **Priya types:** *"Set up my week. Where should I start?"* **The Persona AI responds:** > **Weekly priority checklist, Priya Mehta** > Territory: West Coast mid-market ยท 47 assigned leads ยท Week of Apr 21 > > **Step 1 of 10: ICP scoring complete** > 47 leads scored across 4 dimensions (Firmographic, Technographic, Intent Signals, Win Pattern Match). > > | **Tier** | **Count** | **Action** | > | --------------- | --------- | ------------------------------------------------------------- | > | ๐Ÿ”ฅ HOT (75+) | 5 | Prioritize above everything else this week. | > | ๐ŸŸก WARM (50โ€“74) | 12 | Reach out this week if time permits, next week at the latest. | > | โ„๏ธ LOW (\< 50) | 30 | Route to segmentation, do not discard without guidance. | > > **Top 5 HOT accounts** > > | **Account** | **Score** | **Top signal** | **Recommended first action** | > | ----------------------- | --------- | -------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------ | > | Cascade Data Systems | 92 | Posted VP of Data Platform role 3 weeks ago. Stack matches (Teradata, Python). Closest closed-won match: MedTech Corp (\$142K ARR, 4-month cycle). | Build a prospect profile and draft a personalized email to the hiring manager. | > | Ridgeline Analytics | 88 | Raised \$28M Series C 6 weeks ago. Job postings mention "real-time analytics at scale." | Research the VP of Engineering on LinkedIn, then build an outreach sequence. | > | Apex Financial Group | 84 | Existing Snowflake customer posting about query performance issues on their engineering blog. | Draft a cold email using the blog post as the hook. | > | Pacific Logistics | 79 | Data engineering team doubled in 6 months. Current stack includes Redshift. | Build a prospect profile to identify the right contact. | > | Westfield Manufacturing | 76 | SAP integration need identified from job postings. Manufacturing vertical, 38% historical close rate. | Research the Head of Data and draft outreach leading with SAP integration. | > > **Remaining 42 leads:** 12 WARM accounts summarised with top signal. 30 LOW accounts ready for segmentation when you finish Tier 1 outreach. > > **Suggested follow-ups:** Show me the full score for Cascade Data Systems ยท Build a prospect profile for the VP of Data Platform at Cascade ยท Start outreach for the top 5 ยท Segment my low-fit leads Priya starts with her top lead. **Priya types:** *"Build a prospect profile for the VP of Data Platform at Cascade Data Systems."* **The Persona AI responds:** > **Prospect profile, Cascade Data Systems** > > **Company snapshot** > > | **Attribute** | **Detail** | > | ------------- | ------------------------------------------ | > | Industry | Data infrastructure / SaaS | > | HQ | Portland, OR | > | Revenue | \$85M (estimated) | > | Employees | 420 | > | Data stack | Teradata (migrating), Python, Apache Spark | > > **Trigger events (last 90 days)** > > * Posted VP of Data Platform role 3 weeks ago, job description mentions "migrate from legacy warehouse to a modern, high-performance analytics engine." > * Hired 4 data engineers in the last 60 days. > * CTO spoke at a Portland data meetup about "the cost of slow queries on product decisions." > > **Tech stack signals** > > | **Technology** | **Signal** | **Actian fit** | > | ----------------- | ------------------------------------------- | ---------------------------------------------------------------- | > | Teradata (legacy) | Job postings explicitly reference migration | ๐ŸŸข HIGH, proven migration path, DataConnect accelerates ETL lift | > | Apache Spark | Engineering blog | ๐ŸŸก MODERATE, Vector AI DB complements Spark as the query layer | > | Python | Job postings | ๐ŸŸข HIGH, native Python SDK, clean integration | > > **Contact profiling** > > * **Target:** VP of Data Platform (role posted, not yet filled, hiring manager is likely the CTO or a Sr. Director of Engineering). > * **Pain hypothesis:** Migrating from Teradata means they need a replacement that delivers query performance at scale without the licensing cost. The CTO's public comments about "slow queries" confirm this is a business-level pain, not just a technical wish. > > **Personalisation hooks** > > 1. **Trigger:** The VP of Data Platform job posting mentions "high-performance analytics engine", this is their own language for what Vector AI DB delivers. > 2. **Pain:** Their CTO publicly called out slow queries as a product blocker. Query latency is not an infrastructure issue for them; it is a revenue issue. > 3. **Proof:** MedTech Corp (similar size, similar Teradata migration) closed at \$142K ARR in 4 months. Sub-500ms query times on 600M rows. > > **Suggested follow-ups:** Draft an outreach email using the CTO's meetup talk as the hook ยท Build a LinkedIn message for the CTO ยท Create a full outreach sequence ยท Save this profile to Salesforce Priya drafts the email in one click, sends the LinkedIn message, and loads the full sequence into Salesforce, all before 9:30 am. The rest of her top 5 follow the same pattern. The results: * 47 leads scored and ranked in under 8 seconds, with the specific signal driving each score. * Top 5 accounts identified with a clear first action for each, grounded in real signals. * Prospect profile built from Salesforce, LinkedIn, and the public web in under 30 seconds, including personalization hooks ready to paste into outreach. * Monday morning paralysis replaced with a structured 10-step plan. ## Skills that power all three personas Each persona in IdeaBoxAI is powered by a set of Skills, AI capabilities that define how the Persona AI handles specific types of requests. The Sales Team Persona AI ships with 32 skills out of the box, split across the three personas. You can trigger a skill in three ways: * Click a **scenario card** on the home screen to pre-fill the prompt and run the skill immediately. * Type your request in plain language. The Persona AI auto-detects the appropriate skill from your message. * Type `/` in the chat bar to open the skill picker and select a skill by name. You do not need to memorize skill names or commands. Type what you need in plain language, and the Persona AI matches your intent to the right skill automatically. ### Account Executive skills The following skills handle deal management, meeting preparation, and pipeline workflows. | **Skill** | **What it does** | | ------------------------ | ------------------------------------------------------------------------------------------------------------------------------------- | | Account Brief Generation | Pulls signals from Salesforce, LinkedIn, job postings, and the web to build a 6-section account brief in under 5 minutes. | | SE Alignment Brief | Creates a pre-meeting brief for the Sales Engineer, account context, capabilities to demo, POC outline, and positioning notes. | | Meeting Coaching | Returns the single most important outcome for an upcoming meeting based on deal stage, with exact words to close on it. | | Objection Prep | Generates a prioritized list of likely objections with probability, underlying concern, and conversational responses. | | Pre-Meeting Checklist | Produces a 3-section checklist, materials to bring, Salesforce hygiene items, and talking points to memorize. | | Post-Meeting Log | Turns a freeform meeting debrief into a structured outcome log, deal status, risk flags, and next actions, auto-logged to Salesforce. | | Email Draft | Drafts any sales email, cold outreach, warm follow-up, post-meeting, or post-discovery, under 150 words with a signal-based hook. | | Deal Scorecard | Scores deal health across 5 dimensions (activity, stakeholders, artifacts, close date, competition) with evidence from Salesforce. | | Deal Next Actions | Generates a prioritized action list split into today and this week, with channel, exact words, and the goal each action achieves. | | Re-engage Email | Drafts a direct re-engagement email for a prospect who has gone quiet, no apology, clear reason to reply. | | Mutual Action Plan | Builds a 5-milestone close plan working backward from the target close date, with owners and deliverables for both sides. | | Salesforce Update | Logs session activity to Salesforce, meeting notes, calls, stage changes, and forecast category. Requires explicit confirmation. | | Forecast Guidance | Recommends a forecast category (Commit, Best Case, Pipeline) with evidence-backed talking points for the pipeline review. | | Call Script | Writes a branching call script with a direct opening, conditional paths, a named key ask, and a committed next step with a date. | | Deal Artifact Checklist | Audits which deal documents exist, which are missing, and which gaps are blocking stage progression. | | PDF Export | Exports any brief, scorecard, or plan as a formatted, single-page PDF with branding and a confidential watermark. | ### Sales Engineer skills The following skills handle technical discovery, POC design, and competitive objection handling. | **Skill** | **What it does** | | ----------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------- | | Discovery Question Set | Generates 15โ€“25 tailored discovery questions adapted to industry, tech stack, and deal stage, each annotated with why to ask it. | | Discovery Note Template | Creates a live note-taking template with questions as headers and blank answer fields, auto-saved to Salesforce on completion. | | Post-Discovery Summary | Extracts technical facts from a freeform call debrief, scores Actian fit, and generates a POC recommendation automatically. | | POC Recommendation | Designs a POC scope, dataset, success metric, duration, champion, and products, built to demonstrate advantage against evaluation criteria. | | Rebuttal Card | Returns a structured response to any technical objection in under 3 seconds, including per-competitor positioning for Snowflake, Databricks, and BigQuery. | | Evidence Assembly | Retrieves specific benchmarks, case studies, and certifications matched to the objection category and the prospect's industry. | | Objection Email | Drafts a post-call follow-up addressing technical objections raised during a session, with supporting evidence as named attachments. | | Objection Tracker | Tracks all open objections across a deal, who raised them, what was shared, and whether each one is resolved. | ### Sales Development Rep skills The following skills handle lead scoring, prospect research, and outbound sequencing. | **Skill** | **What it does** | | ------------------------- | ------------------------------------------------------------------------------------------------------------------------------- | | Prospect Profile | Builds an enriched profile, company snapshot, trigger events, tech stack signals, pain hypothesis, and 3 personalization hooks. | | LinkedIn Message | Writes a 3-sentence LinkedIn message with a signal-based hook, a value sentence, and a low-friction question as the CTA. | | Outreach Sequence | Builds a 5-touch, 14-day outreach sequence across email, LinkedIn, and phone, each touch anchored to a real signal. | | ICP Score (Batch) | Scores an entire lead list against the ICP model and ranks every account by conversion likelihood with heat labels. | | ICP Score (Single) | Gives a full ICP breakdown on one account, every dimension scored, with the closest closed-won match named as reference. | | Low-Fit Segmentation | Segments low-scoring leads into Nurture, Disqualify, and Needs More Data, with specific next steps for each group. | | Weekly Priority Checklist | Sets up the week with a 10-step prioritization routine, from scoring leads to drafting emails to loading sequences. | | Salesforce Lead Sync | Saves outbound research, email drafts, trigger events, and ICP scores to Salesforce. Requires explicit confirmation. | ## Connected to your sales data The Sales Team Persona AI is grounded in your live data, not a sample, not a simulation. It connects to the systems your sales team already uses, pulling information in real time to inform every response. When an AE asks "prep me for my meeting with Meridian Health," the Persona AI queries Salesforce for opportunity data, searches LinkedIn for recent activity, scans job postings for tech stack signals, and pulls prior notes from Confluence, all in parallel, in under 5 minutes. Each persona only sees the data relevant to their role. An AE sees opportunities, contacts, and deal activity. An SE sees technical notes, discovery records, and POC status. An SDR sees leads, ICP scores, and outreach history. The platform connects through the following layers: * Salesforce for CRM data, opportunities, contacts, leads, activity history, and forecast categories. * Confluence for internal knowledge, battlecards, case studies, prior SE notes, and RFP templates. * Google Drive for deal documents, proposals, account briefs, and mutual action plans. * LinkedIn for prospect research, role changes, company signals, and professional context. * Public web for company intelligence, news, job postings, financial filings, and tech stack signals. ## Getting started Setting up the Sales Team Persona AI for your team takes less than a day. Link Salesforce, Confluence, Google Drive, and LinkedIn from the Connections settings in the Admin Console. Each integration is configured per-organization and tested to confirm data visibility. Set up Account Executive, Sales Engineer, and SDR personas in the Admin Console. Assign the relevant skills and knowledge base to each. The knowledge base scopes each persona to their relevant data sources automatically. Click the knowledge base generation button. The platform maps your Salesforce object structure, indexes your Confluence content, and generates the cubes needed for Agent BI dashboards. This takes under 10 minutes. Add your sales team members and assign them to their persona. They log in and see their role-specific Persona AI immediately. Scenario cards guide them through the most common workflows from day one. Start on the IdeaBoxAI dev environment to validate connections and test persona responses against your real Salesforce data. Once the team is confident, promote to production. The Sales Team Persona AI is available now as part of IdeaBoxAI's Persona AI suite. Contact the IdeaBoxAI team to set up a live demo connected to your Salesforce environment. ## Next steps Explore these guides to learn more about the platform capabilities behind the Sales Team Persona AI. Create and configure role-specific personas in the Admin Console. Assign skills to personas and customize how the Persona AI handles requests. Connect structured data sources and generate knowledge bases for grounded responses. Understand the measurable outcomes the Conversational AI delivers. # Driving business outcomes Source: https://docs.ideaboxai.com/copilot/use-cases/business-outcome Understand the business problems the Persona AI solves and the measurable outcomes it delivers across sales teams. The Persona AI is designed to address specific, measurable inefficiencies across sales workflows. This page maps each problem area to the platform capabilities that solve it. ## Problems addressed The following table maps each problem area to its current state and the business impact it creates. | Problem Area | Current State | Business Impact | | ------------------------------ | --------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------ | | **Manual research overhead** | AEs spend 30-60 min per account on research with no standard format. | Inconsistent discovery, missed signals, lost selling time. | | **Invisible deal risk** | No systematic tracking of missing artefacts, stale activity, or single-threaded deals. | Surprise slips, forecast inaccuracy, avoidable losses. | | **Generic outreach and demos** | SDRs reuse stale templates; SEs configure demos manually without deal context. | Low reply rates, poor demo-to-POC conversion. | | **Fragmented knowledge** | Battlecards, RFP answers, and objection responses scattered across multiple systems and tribal knowledge. | Inconsistent positioning, slow responses, long SE ramp time. | ## How the Persona AI solves each problem * **Research in under 5 minutes**: The Account Research skill generates structured value briefs, company snapshot, strategic priorities, tech stack, pain-to-product mapping, and talk tracks, in under 5 minutes instead of 30-60 minutes. * **Proactive deal risk detection**: The Deal Health Scorecard flags missing artifacts, stale activity, single-threaded contacts, and incomplete qualification fields. Momentum alerts fire when deal health degrades. * **Personalized outreach at scale**: The Outreach Email Drafting skill generates AI-personalised emails with A/B variants tailored to prospect role, industry, and intent signals, in under 2 minutes per prospect. * **Centralized knowledge access**: The Persona AI queries Confluence, Google Drive, and prior submissions to draft RFP responses, surface battlecards, and handle objections with consistent positioning. ## How the Persona AI helps you get things done faster The Persona AI accelerates core tasks by eliminating manual data gathering and document assembly. The following examples show the time savings across common activities. | Task | Without Persona AI | With Persona AI | | ------------------------ | ---------------------------------------------------- | ---------------------------------------------------------- | | Account brief generation | 30-60 min of manual research across multiple systems | Under 5 min, auto-generated from connected data | | Proposal creation | 2-4 hours of compiling data and formatting | Under 30 min, generated from templates and live data | | Prospect research | Variable, manual lookups across tools | Under 2 min per prospect | | RFP response drafting | Days of searching prior submissions and writing | 60%+ faster with AI-assisted drafting from knowledge bases | | Deal risk assessment | Periodic manual review, often too late | Continuous, with proactive alerts on degrading deal health | | Follow-up email drafting | 15-30 min per personalised email | Under 2 min with context from CRM and call data | The Persona AI removes the repetitive data gathering that consumes most of the working day, so your team spends time on decisions and relationships instead of copying between systems. # Mortgage Insurance Source: https://docs.ideaboxai.com/copilot/use-cases/enact-mortgage-insurance-case-study Learn how IdeaBoxAI's Mortgage Insurance Persona AI transforms data discovery, underwriting decisions, and customer support for a private mortgage insurer, connecting Actian DI, Snowflake, and Google Drive into a single persona-aware assistant. Every ML engineer at a mortgage insurer knows the dataset hunt. You need to build a credit scoring model. There are 47 tables in Snowflake that look relevant. Three of them have "customer" in the name. One is certified, one is a staging copy, and one is a raw extract that nobody owns. The metadata catalog tells you the tables exist, but not which one to trust, whether it is current, or what is missing for your use case. You spend two days figuring out which data you can actually use before writing a single line of model code. Every underwriter knows the tab-switching. A mortgage application is on your desk. You need the applicant's risk factor from Snowflake, the LTV coverage rules from the policy guide, and the pipeline status to see how many other applications are waiting. Three systems, three logins, three different interfaces, just to make one approval decision. Every customer support agent knows the hold time. A borrower calls asking about their coverage. You need to look up their policy, check their claim status, and confirm the cancellation rules, all while the customer waits. If the answer is not in the first document you open, you put them on hold and start searching. The Mortgage Insurance Persona AI solves all three. It connects Actian DI (Zeenea) for metadata intelligence, Snowflake for live structured data, and Google Drive for policy documentation, unified into a single persona-aware assistant accessible from the browser and Microsoft Teams. ## The knows your role before you ask A Persona in IdeaBoxAI is the AI's understanding of who you are, what you need, and what your data looks like. When you log in as an ML Engineer, you get a data intelligence layer tuned to dataset discovery, certification status, and lineage. When you log in as an Underwriter, you get a decisioning layer tuned to application pipelines, risk factors, and policy rules. When you log in as a Customer Support Agent, you get a resolution layer tuned to coverage questions, claim status, and escalation summaries. The Mortgage Insurance Persona AI ships with three purpose-built personas. Each connects to the data sources relevant to their role. The persona you are assigned to determines which data you see, which skills are available, and which scenario cards appear on your home screen. The following table summarises the three personas. | **Persona** | **Focus** | **Key Needs** | | ---------------------- | ------------------------------------ | ---------------------------------------------------------------------------------------------------------- | | Data and ML Engineer | Datasets, models, data quality | Certified dataset discovery, table disambiguation, model gap analysis, lineage tracking, NL-to-SQL queries | | Underwriter | Applications, risk, policy rules | Application pipeline visibility, risk filtering, eligibility decisions, policy lookups, pipeline aging | | Customer Support Agent | Policies, claims, borrower questions | Coverage answers, claim status, cancellation rules, escalation summaries | Each persona also comes with pre-configured scenario cards on the home screen. Click any card to pre-fill the chat input with a common workflow prompt. You can also type your request directly in plain language; the Persona AI auto-detects the right skill. To browse all available skills, type `/` in the chat bar to open the skill picker. ## ML Engineer scenario: dataset discovery for model building **Anika Patel ยท Senior ML Engineer** Credit risk team ยท Building a new credit scoring model ยท Needs certified training data Anika has been asked to build a new credit scoring model. Before she can start, she needs to find which datasets are available, which ones are certified for production use, and whether any required features are missing. Before IdeaBoxAI, this meant opening the Atacama catalog, searching for relevant tables, then cross-referencing Snowflake to check freshness and row counts, a process that usually took 1โ€“2 days. **Anika types:** *"Which datasets are certified and trusted for building a credit scoring model?"* **The Persona AI responds:** > **Certified dataset discovery, credit scoring** > 47 tables in Snowflake match credit-related metadata ยท 8 are certified for production use ยท 3 are directly relevant to credit scoring > > **Certified datasets for credit scoring** > > | **Table** | **Schema** | **Certification** | **Owner** | **Last updated** | **Row count** | > | -------------------------- | ----------------- | ----------------- | ------------------- | ---------------------------- | ------------- | > | CREDIT\_SCORE\_FEATURES | PROD.RISK | โœ… Certified | Risk Analytics Team | Apr 22, 2026 (daily refresh) | 2.4M | > | LOAN\_APPLICATION\_HISTORY | PROD.UNDERWRITING | โœ… Certified | Underwriting Ops | Apr 22, 2026 (daily refresh) | 8.1M | > | BORROWER\_PROFILE | PROD.CUSTOMER | โœ… Certified | Data Engineering | Apr 21, 2026 (daily refresh) | 3.2M | > > **Why these three:** > > * CREDIT\_SCORE\_FEATURES contains the core feature set used in the current production model, FICO scores, DTI ratios, employment history, and 14 derived risk indicators. > * LOAN\_APPLICATION\_HISTORY provides outcome labels (approved, denied, defaulted) across 8.1M historical applications. Essential for supervised training. > * BORROWER\_PROFILE links applicant demographics and LTV ratios to each application record. > > **Potentially relevant but not certified** > > | **Table** | **Schema** | **Issue** | **Recommendation** | > | -------------------------- | -------------- | -------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | > | RAW\_CREDIT\_BUREAU | STAGING.RISK | Not certified, raw extract from credit bureau feed | Do not use for model training. This is the upstream source for CREDIT\_SCORE\_FEATURES. Use the certified downstream table instead. | > | CUSTOMER\_PAYMENT\_HISTORY | PROD.SERVICING | Certified for servicing, not for credit scoring | Contains 36 months of payment behaviour. Relevant for churn prediction but not labelled for credit scoring. Request cross-certification from the Risk Analytics team if you need it. | > > **Data lineage note:** CREDIT\_SCORE\_FEATURES is derived from RAW\_CREDIT\_BUREAU through 3 transformation stages. The most recent pipeline run completed at 2:14 am today. No quality alerts in the last 30 days. > > **Suggested follow-ups:** Show me the full column list for CREDIT\_SCORE\_FEATURES ยท What features are missing for a churn prediction model ยท Trace the lineage for BORROWER\_PROFILE ยท Run a sample query on LOAN\_APPLICATION\_HISTORY Anika follows up with a question that used to require opening Slack and asking the data engineering team. **Anika types:** *"Is BANK\_CUSTOMER the authoritative table or should I use RAW\_CUSTOMER or STAGING\_CUSTOMER?"* **The Persona AI responds:** > **Table disambiguation, CUSTOMER tables** > 3 tables match "customer" ยท 1 is authoritative > > | **Table** | **Schema** | **Status** | **Refresh** | **Row count** | **Verdict** | > | ----------------- | ---------------- | ------------------------------- | ------------- | ------------- | --------------------------------------------------------------------------------------------------------------------------------------------------- | > | BANK\_CUSTOMER | PROD.CUSTOMER | โœ… Certified ยท Authoritative | Daily, 3:00am | 3.2M | **Use this one.** This is the certified production table. Owned by Data Engineering. | > | STAGING\_CUSTOMER | STAGING.CUSTOMER | โš ๏ธ Not certified ยท Staging copy | Hourly | 3.2M | Do not use. This is the staging table that feeds BANK\_CUSTOMER. Data may include incomplete records from the current day's ingest. | > | RAW\_CUSTOMER | RAW\.CUSTOMER | โŒ Not certified ยท Raw extract | On ingest | 4.1M | Do not use. Raw extract from the source system. Contains 900K records that failed validation and were filtered out in the staging-to-prod pipeline. | > > **Key difference:** RAW\_CUSTOMER has 4.1M rows vs BANK\_CUSTOMER's 3.2M. The 900K gap is intentional, those records failed data quality checks (missing SSN, invalid state codes, or duplicate applicant IDs). If your model trains on RAW\_CUSTOMER, it will include invalid records that the production pipeline already rejected. > > **Lineage:** Source system โ†’ RAW\_CUSTOMER โ†’ STAGING\_CUSTOMER โ†’ BANK\_CUSTOMER. The BANK\_CUSTOMER table is the end of the validated pipeline. Training a model on uncertified tables risks embedding data quality issues into production predictions. The Persona AI surfaces certification status and lineage so engineers can make informed decisions, not assumptions. The results: * Certified datasets identified in seconds, with ownership, freshness, and row counts from the live metadata catalog. * Table disambiguation resolved without asking the data engineering team on Slack. The 900K record gap between raw and production tables explained with specific validation reasons. * 1โ€“2 days of dataset hunting reduced to a 3-minute conversation. ## Underwriter scenario: application review and eligibility **James Wu ยท Senior Underwriter** Reviews 15โ€“20 mortgage applications per day ยท Pipeline review with manager at 11 am James reviews mortgage applications for approval. Each decision requires checking the applicant's risk profile in Snowflake, confirming eligibility against policy rules in the coverage guide, and tracking how long the application has been in review. Before IdeaBoxAI, this meant three browser tabs and a spreadsheet to track aging. **James types:** *"Show me all pending mortgage applications waiting for my approval."* **The Persona AI responds:** > **Application pipeline, James Wu** > 18 applications pending ยท 3 flagged for attention > > **๐Ÿ”ด Priority, aging or high risk** > > | **Application** | **Applicant** | **Loan amount** | **LTV** | **Credit score** | **Risk factor** | **Days in review** | **Flag** | > | --------------- | --------------- | --------------- | ------- | ---------------- | --------------- | ------------------ | ---------------------------------------------------------------------------- | > | APP-28471 | Martinez, Elena | \$340,000 | 92% | 618 | 14.2 | **8 days** | Aging, exceeds 5-day SLA. Risk factor above threshold. | > | APP-28455 | Chen, Robert | \$520,000 | 88% | 642 | 12.8 | **6 days** | Aging, exceeds 5-day SLA. | > | APP-28502 | Johnson, Tanya | \$275,000 | 95% | 580 | 16.1 | 2 days | Highest risk factor in queue. Requires senior review per policy section 4.2. | > > **๐ŸŸก Standard review, 12 applications** > All within SLA. Risk factors between 4.0 and 10.0. Average days in review: 2.4. > > | **Risk range** | **Count** | **Average loan amount** | **Average LTV** | > | ------------------- | --------- | ----------------------- | --------------- | > | 4.0โ€“6.0 (low risk) | 5 | \$310,000 | 78% | > | 6.1โ€“8.0 (moderate) | 4 | \$385,000 | 84% | > | 8.1โ€“10.0 (elevated) | 3 | \$420,000 | 89% | > > **๐ŸŸข Ready for fast-track, 3 applications** > All three have risk factors below 4.0, LTV under 80%, and credit scores above 740. These qualify for automated approval under policy section 2.1. > > **Suggested follow-ups:** Show me the full details for APP-28471 ยท Check eligibility for APP-28502 ยท How many applications this week have a risk factor above 12 ยท Show me my pipeline aging trend over the last 30 days James clicks into the highest-risk application. **James types:** *"I have an applicant with a 580 credit score and 10% down, are they eligible?"* **The Persona AI responds:** > **Eligibility decision, 580 credit score, 10% down payment** > > | **Criteria** | **Applicant** | **Policy requirement** | **Status** | > | -------------- | ------------- | ---------------------------------- | --------------------- | > | Credit score | 580 | Minimum 620 for standard approval | โŒ Below threshold | > | Down payment | 10% (LTV 90%) | Minimum 5% required | โœ… Meets requirement | > | LTV ratio | 90% | Maximum 97% for insured loans | โœ… Within limit | > | Debt-to-income | Not provided | Maximum 45% per policy section 3.4 | โš ๏ธ Needs verification | > > **Eligibility verdict: Not eligible for standard approval.** > > The 580 credit score is 40 points below the minimum threshold of 620 defined in policy section 3.1 (Minimum Credit Requirements). This is a hard floor, it cannot be waived at the underwriter level. > > **Alternative paths:** > > * If the applicant has compensating factors (12+ months of reserves, no derogatory marks in 24 months), the application may qualify for exception review under policy section 3.1.3. Exception reviews require VP-level sign-off. > * If the credit score is within 90 days of a rescore event (dispute resolution, paid collection), recommend the applicant reapply after the rescore. The current application should be placed on hold, not denied. > > **Policy source:** Coverage Guide, Section 3.1, Minimum Credit Requirements. Last updated: January 2026. > > **Suggested follow-ups:** Show me the full policy on exception reviews ยท Check if this applicant has compensating factors ยท What is the maximum LTV ratio for a 620 credit score ยท Pull the DTI requirements for this loan amount The Persona AI pulls eligibility rules from the policy guide in Google Drive and cross-references applicant data from Snowflake. Every answer cites the specific policy section so the underwriter can verify the source. The results: * 18 applications surfaced with risk scores, aging status, and SLA flags, no spreadsheet tracking needed. * Eligibility decision returned in seconds with the specific policy section, threshold, and alternative paths. * 3 applications auto-identified for fast-track approval, freeing underwriter time for the complex cases. * Pipeline aging visible at a glance, with applications exceeding the 5-day SLA flagged automatically. ## Underwriter scenario: risk filtering before pipeline review **James Wu ยท Senior Underwriter** Pipeline review with manager in 30 minutes ยท Needs to know which applications carry the highest risk James needs to walk into his pipeline review with a clear picture of risk concentration. His manager will ask how many high-risk applications are in the queue and what the team is doing about them. **James types:** *"How many applications this week have a RISK\_FACTOR above 12?"* **The Persona AI responds:** > **Risk filtering, RISK\_FACTOR above 12** > Week of April 21, 2026 ยท 74 applications received ยท 6 have a risk factor above 12 > > | **Application** | **Applicant** | **Risk factor** | **Credit score** | **LTV** | **Loan amount** | **Days in review** | **Assigned to** | > | --------------- | --------------- | --------------- | ---------------- | ------- | --------------- | ------------------ | --------------- | > | APP-28502 | Johnson, Tanya | 16.1 | 580 | 95% | \$275,000 | 2 | James Wu | > | APP-28471 | Martinez, Elena | 14.2 | 618 | 92% | \$340,000 | 8 | James Wu | > | APP-28519 | Okafor, David | 13.8 | 625 | 91% | \$410,000 | 1 | Sarah Lin | > | APP-28487 | Patel, Neha | 13.1 | 630 | 88% | \$295,000 | 4 | James Wu | > | APP-28510 | Reeves, Mark | 12.6 | 640 | 90% | \$380,000 | 2 | Sarah Lin | > | APP-28455 | Chen, Robert | 12.8 | 642 | 88% | \$520,000 | 6 | James Wu | > > **Risk concentration analysis:** > > * 6 of 74 applications this week (8.1%) are above the 12.0 threshold. This is slightly above the trailing 4-week average of 6.8%. > * 4 of the 6 are assigned to James Wu. Consider redistributing APP-28487 or APP-28455 to balance the high-risk review load. > * The common pattern: all 6 have LTV above 88% and credit scores below 645. The risk is driven by thin equity combined with borderline credit, not a single outlier factor. > > **SLA status:** 2 of the 6 (APP-28471 and APP-28455) have exceeded the 5-day review SLA. Both are assigned to James. > > **Suggested follow-ups:** Show me the risk factor trend over the last 4 weeks ยท Check eligibility for APP-28502 ยท Redistribute my high-risk applications ยท Compare this week's risk profile to the same week last quarter James walks into his review with the exact numbers: 6 high-risk applications, 8.1% concentration (above the 4-week average), two past SLA, and a clear pattern linking thin equity to borderline credit. ## Support scenario: coverage question and claim lookup **Diana Reyes ยท Customer Support Agent** Handles 40โ€“50 borrower calls per day ยท Current call: borrower asking about coverage Diana is on a call with a borrower who wants to know what their mortgage insurance covers and whether they can cancel it early. Before IdeaBoxAI, she would put the borrower on hold, open the coverage guide PDF, search for the relevant section, and hope the search term matched. **Diana types:** *"What is the coverage for a homebuyer who put 15% down?"* **The Persona AI responds:** > **Coverage lookup, 15% down payment** > > A homebuyer who put 15% down has a loan-to-value (LTV) ratio of 85%. The following coverage applies. > > | **Coverage attribute** | **Detail** | > | ------------------------------- | ----------------------------------------------------- | > | LTV at origination | 85% | > | MI coverage percentage | 12% of the loan amount | > | Monthly premium (typical range) | 0.30%โ€“0.55% of the loan amount annually, paid monthly | > | Coverage type | Borrower-paid mortgage insurance (BPMI) | > > **What the coverage means in practice:** > If the borrower defaults and the lender files a claim, mortgage insurance covers 12% of the original loan amount. On a $300,000 loan, that is $36,000 in coverage to the lender. > > **Cancellation eligibility:** > > * The borrower can request cancellation when LTV reaches 80% (through principal paydown or home appreciation). This requires a current appraisal. > * Automatic cancellation occurs when LTV reaches 78% based on the original amortization schedule, per the Homeowners Protection Act. > > **Policy source:** Coverage Guide, Section 2.3, Standard Coverage Percentages by LTV Band. Last updated: January 2026. > > **Suggested follow-ups:** At what LTV ratio does MI automatically cancel ยท Can a borrower cancel early with a new appraisal ยท What is the coverage for a 10% down payment ยท Look up a specific policy number The borrower asks a follow-up. Diana does not need to put them on hold. **Diana types:** *"Can a borrower cancel their mortgage insurance early and if so, how?"* **The Persona AI responds:** > **Cancellation rules, early cancellation** > > A borrower can cancel mortgage insurance early through two paths. > > **Path 1: Borrower-requested cancellation (LTV reaches 80%)** > > | **Requirement** | **Detail** | > | --------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------- | > | LTV threshold | 80% or below | > | How LTV is calculated | Based on current principal balance vs original property value, or a new appraised value if the borrower requests a reappraisal | > | Payment history | Must be current on payments. No 30-day late payments in the last 12 months. No 60-day late payments in the last 24 months. | > | Process | Borrower submits a written request to the servicer. Servicer orders appraisal if needed. Cancellation takes effect on the first day of the month following approval. | > > **Path 2: Automatic cancellation (LTV reaches 78%)** > > | **Requirement** | **Detail** | > | ------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------- | > | LTV threshold | 78% based on the original amortisation schedule | > | No borrower action needed | The servicer is required to cancel MI automatically on the date the principal balance is scheduled to reach 78% of the original property value. | > | Legal basis | Homeowners Protection Act of 1998 (HPA) | > > **Key distinction:** Borrower-requested cancellation at 80% can use a new appraisal to account for home value appreciation. Automatic cancellation at 78% uses only the original amortization schedule; home appreciation is not factored in. > > **What to tell the borrower:** "You can request cancellation now if your balance is at or below 80% of your home's current value. If you believe your home has appreciated, you can request a reappraisal through your servicer. Otherwise, your MI will automatically cancel when your scheduled balance reaches 78% of the original value, no action needed on your end." > > **Policy source:** Coverage Guide, Section 6.1, Cancellation and Termination. Homeowners Protection Act of 1998. > > **Suggested follow-ups:** Look up a specific borrower's current LTV ยท When will MI auto-cancel for a specific loan number ยท What is the reappraisal process ยท Explain the difference between BPMI and LPMI Diana answers the borrower's question in real time, no hold, no document search, no transfer to a senior agent. **Diana types:** *"What is the current status of claim number 48291?"* **The Persona AI responds:** > **Claim status, #48291** > > | **Field** | **Detail** | > | ------------------- | ------------------------------------------------------- | > | Claim number | 48291 | > | Policy number | PMI-2024-07-38291 | > | Borrower | Williams, Karen | > | Loan amount | \$285,000 | > | Claim filed | March 14, 2026 | > | Claim type | Default, 90+ days delinquent | > | Status | **Under review**: assigned to Claims Analyst Team B | > | Documents received | 3 of 4 required (missing: updated property valuation) | > | Expected resolution | Within 15 business days of complete document submission | > > **What is holding this claim:** The updated property valuation has not been received. This document was requested from the servicer on March 18. It has been 26 business days with no submission. > > **Recommended action:** Contact the servicer to follow up on the outstanding property valuation. Once received, the claim moves to final review and should resolve within 15 business days. > > **Suggested follow-ups:** Draft a follow-up email to the servicer about the missing valuation ยท Show me all claims for this borrower ยท Summarise this claim for escalation to a senior agent The Persona AI pulls claim data from Snowflake and policy details from Google Drive in a single query. Support agents never need to switch systems, the answer arrives with the source cited. The results: * Coverage question answered in seconds with the specific policy section, premium range, and cancellation rules, no PDF searching. * Borrower-facing language provided alongside the technical answer so the agent can respond naturally. * Claim status pulled with the specific blocker identified (missing document, 26 days outstanding) and the next action ready. * Average handle time reduced by eliminating hold time, system switching, and manual document lookup. ## Skills that power all three personas Each persona in IdeaBoxAI is powered by a set of Skills, AI capabilities that define how the Persona AI handles specific types of requests. The Mortgage Insurance Persona AI ships with 15 skills across the three personas. You can trigger a skill in three ways: * Click a **scenario card** on the home screen to pre-fill the prompt and run the skill immediately. * Type your request in plain language. The Persona AI auto-detects the appropriate skill from your message. * Type `/` in the chat bar to open the skill picker and select a skill by name. You do not need to memorize skill names or commands. Type what you need in plain language and the Persona AI matches your intent to the right skill automatically. ### Data and ML Engineer skills The following skills handle dataset discovery, data quality validation, and analytical queries against the warehouse. | **Skill** | **What it does** | | --------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Dataset Discovery | Searches the Actian DI (Zeenea) catalog for certified, trusted datasets that match a use case. Returns certification status, owner, freshness, and row counts. | | Table Disambiguation | Compares similarly named tables and identifies the authoritative source. Surfaces lineage, certification, and the differences between raw, staging, and production copies. | | Model Gap Analysis | Maps available datasets against the requirements for a specific model type (credit scoring, churn, risk). Identifies what exists, what is missing, and who owns each gap. | | NL-to-SQL | Translates natural language questions into SQL queries against Snowflake. Returns results directly, bypassing the catalog for live structured queries. | | Lineage and Ownership | Traces a table's full data lineage from source system to production. Shows transformation stages, refresh schedules, and the team responsible for each step. | ### Underwriter skills The following skills handle application reviews, risk assessment, and policy-based eligibility decisions. | **Skill** | **What it does** | | -------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Application Pipeline | Surfaces all pending mortgage applications assigned to the underwriter. Flags aging, risk level, and SLA status per application. | | Risk Filtering | Queries Snowflake for applications matching risk criteria, risk factor thresholds, credit score ranges, or LTV bands. Returns counts, distributions, and trends. | | Policy Q\&A | Answers policy-related questions by querying the coverage guide in Google Drive. Returns the answer with the specific section cited. | | Eligibility Decision | Evaluates an applicant's profile against policy eligibility criteria. Returns a pass/fail verdict per criterion with the specific policy threshold and alternative paths. | | Pipeline Aging | Identifies applications that have exceeded review SLAs. Ranks by days outstanding and flags patterns in aging concentration. | ### Customer Support Agent skills The following skills handle borrower-facing questions, claim lookups, and escalation preparation. | **Skill** | **What it does** | | ------------------ | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Coverage Q\&A | Answers coverage questions based on loan characteristics (LTV, down payment, loan type). Returns coverage percentages, premium ranges, and cancellation eligibility. | | Policy Q\&A | Answers policy and regulatory questions from the coverage guide. Returns the answer with borrower-facing language and the source section cited. | | Claim Status | Looks up a claim by number or borrower name. Returns the current status, assigned team, outstanding documents, and expected resolution timeline. | | Cancellation Rules | Explains the paths to mortgage insurance cancellation, borrower-requested and automatic, with requirements, process steps, and legal basis. | | Escalation Summary | Assembles a structured summary of an account dispute or complex case, pulling data from Snowflake and policy context from Google Drive, ready to hand off to a senior agent. | ## Connected to your data infrastructure The Mortgage Insurance Persona AI sits on top of three data sources, unified into a single persona-aware assistant. It does not replace your existing infrastructure, it makes it accessible through natural language. When an ML engineer asks "which datasets are certified for credit scoring?", the Persona AI queries the Actian DI (Zeenea) catalog for certification status and lineage, then cross-references Snowflake for freshness and row counts. The answer arrives in seconds, with the specific tables, their owners, and the reasons to use or avoid each one. Each persona only sees the data relevant to their role. An ML engineer queries the metadata catalog and Snowflake. An underwriter queries Snowflake application data and Google Drive policy documents. A support agent queries Snowflake claim records and Google Drive coverage guides. The platform connects through the following layers: * Actian DI (Zeenea) connected via API, loads metadata context, data lineage, and certification status from the enterprise catalog. * Snowflake connected as a structured knowledge base, NL-to-SQL queries run directly against the warehouse for live application, claim, and risk data. * Google Drive connected via RAG pipeline, retrieval-augmented Q\&A over policy documents, coverage guides, and underwriting manuals. * IdeaBoxAI browser app for all personas, dedicated web interface with full conversational and dashboard capabilities. * Microsoft Teams bot integration, Teams-native queries for users who work inside Teams throughout the day. ## Getting started Setting up the Mortgage Insurance Persona AI for your team takes less than a day. Link Actian DI (Zeenea) via API, Snowflake as a structured knowledge base, and Google Drive as a document source from the Connections settings in the Admin Console. Each integration is configured and tested to confirm data visibility. Set up Data and ML Engineer, Underwriter, and Customer Support Agent personas in the Admin Console. Assign the relevant skills and knowledge base to each. The knowledge base scopes each persona to their relevant data sources automatically. Click the knowledge base generation button. The platform maps your Snowflake schema, indexes your Google Drive policy documents, and connects to the Actian DI catalog. This takes under 10 minutes. Add your team members and assign them to their persona. They log in and see their role-specific Persona AI immediately. Scenario cards guide them through the most common workflows from day one. Start on the IdeaBoxAI dev environment to validate connections and test persona responses against your real data. Once the team is confident, promote to production. The Mortgage Insurance Persona AI is available now as part of IdeaBoxAI's Persona AI suite. Contact the IdeaBoxAI team to set up a live demo connected to your data infrastructure. ## Next steps Explore these guides to learn more about the platform capabilities behind the Mortgage Insurance Persona AI. Create and configure role-specific personas in the Admin Console. Assign skills to personas and customize how the Persona AI handles requests. Connect structured data sources and generate knowledge bases for grounded responses. Understand the measurable outcomes the Conversational AI delivers. # Construction & Real Estate Source: https://docs.ideaboxai.com/copilot/use-cases/sage-cre-copilot-case-study Learn how IdeaBoxAI's Construction & Real Estate Persona AI transforms AP, project management, and field engineering workflows in construction and real estate, powered by 24 purpose-built skills across three personas. Every AP Coordinator knows the Friday morning feeling. Forty-seven open invoices. A cash position that won't cover all of them. Three vendors calling about payment status. A lien deadline on Monday that nobody flagged until now. Every Project Manager knows the equivalent. A client wants an owner report by 3 pm. You have nine active jobs. Three of them are probably over budget. You just don't know which three yet, or by how much, or why. Every Project Engineer knows the daily grind. Five cost codes are burning faster than schedule percentage. Two RFIs are overdue with no response. A subcontractor's insurance expired last week, and nobody noticed. The Sage Persona AI solves all three, not with another report or another dashboard, but with a Persona AI that connects directly to your Sage 300 CRE data via ODBC. It understands your role and answers your actual questions in plain English. ## Getting started Setting up the Sage Persona AI for your team takes less than a day. Link your Sage environment from the Connections settings in the platform. The ODBC connector is configured with a read-only service account on your Sage 300 server network. No data leaves your network; only summarised query results are sent to the AI layer. Set up Project Manager, Project Engineer, and AP Coordinator personas in the Admin Console. Assign the relevant skills and knowledge base to each. The knowledge base scopes each persona to their relevant Sage tables automatically. Click the knowledge base generation button. The platform uses AI to map your Sage table structure, understand the relationships between modules, and generate the cubes needed for dashboards. This takes under 10 minutes. Add your team members and assign them to their persona. They log in and see their role-specific Persona AI and dashboard immediately. Starter prompts guide them through the most common tasks from day one. Start on the IdeaBoxAI dev environment to validate the connection and test the persona responses against your real data. Once the team is satisfied, promote to production. The Sage Persona AI is available now as part of IdeaBoxAI's Persona AI suite. Contact the IdeaBoxAI team to set up a live demo connected to your Sage 300 CRE environment. ## The Persona AI knows your role before you ask A Persona in IdeaBoxAI is the AI's understanding of who you are, what you need, and what your data looks like. The persona you are assigned to determines which data you see, which skills are available, and which starter prompts appear on your home screen. The Sage Persona AI ships with three purpose-built personas. Each connects to the Sage 300 CRE tables relevant to their role via ODBC. | **Persona** | **Focus** | **Key Needs** | | ---------------- | --------------------------------------------------- | --------------------------------------------------------------------- | | Project Manager | Portfolio P\&L, owner reporting, change orders | Budget vs actual, profit forecast, schedule risk, cash flow | | Project Engineer | Job cost by cost code, RFIs, subcontract compliance | Cost code burn rates, overdue RFIs, change order drafting, daily logs | | AP Coordinator | Invoices, payments, vendor compliance | PO matching, payment prioritisation, W-9 and insurance tracking | ## AP scenario: payment run prioritization **Priya Sharma - AP Coordinator** 12 active projects, 47 open invoices, cash position review due by 9 am. Priya types: *"Which open commitment invoices should I pay first this week, and are any of them at lien risk?"* **The Persona AI responds** with a payment priority summary covering all 47 invoices ranked by lien risk, early-pay discount expiry, due date, and PO match status: > **Payment priority summary, week of April 17:** > 47 open invoices across 12 projects. \$1.24M total outstanding. Cash available: \$940,000. > > **Action required today:** > > * **ABC Electrical, INV-2301** (\$84,200): Lien deadline Mon Apr 21. No waiver on file. Preliminary notice filed Feb 3. > * **XYZ Mechanical, INV-2289** (\$12,400): 2% early-pay discount expires today at 5 pm. PO matched, no holds. > * **ABC Plumbing, INV-2294** (\$31,000): Lien deadline Thu Apr 24. Waiver requested Mar 28, no response in 20 days. > > **Recommended payment run:** 6 invoices totaling \$183,200. All have approved POs, no compliance holds. > > **Cash position after full run:** \$629,200 remaining. > **38 remaining invoices:** \$929,200 total. None due before Apr 28. Priya follows up: *"Show me all invoices from ABC Electrical across all jobs."* The Persona AI returns a vendor summary with 4 invoices across 2 projects, identifies \$115,600 in lien-exposed invoices with no waivers on file, and recommends requesting conditional lien waivers before releasing payment. **The results:** * 47 invoices ranked by risk in under 2 minutes. * \$115,600 in lien-exposed invoices flagged with vendor history. * Friday morning prep reduced from 90 minutes to a 5-minute review. ## AP scenario: daily morning brief **Marcus Okafor - AP Coordinator** 8 active construction projects, 3 states, 31 open invoices Marcus types: *"Give me my AP morning brief, what needs my attention today?"* **The Persona AI responds** with a structured brief covering action items, payments due, and a week-ahead outlook: > **AP morning brief, Thursday, April 17, 2026** > 8 active projects. 31 open invoices. \$1.24M total AP liability > > **Action required today:** > > * XYZ HVAC, INV-4410 (\$18,400, Midtown Tower): No matching PO found. First mismatch for this vendor across 6 prior invoices. > * ABC Framing: Insurance cert expires Mon Apr 21. Active on 2 projects. Send renewal request today. > > **Due in 48 hours, ready to process:** > 3 invoices totaling \$62,800. All PO matched, no holds, waivers on file. > > **Heads up, next 7 days:** > XYZ MEP insurance expires in 11 days. 4 invoices totaling \$87,400 due between Apr 22-25. Marcus reviews the brief in under 3 minutes, flags XYZ HVAC for investigation, and approves the payment run before his 9 am stand-up. ## PM scenario: portfolio health scan **Sarah Chen - Project Manager** 9 active jobs, portfolio review with VP at 10 am Sarah types: *"Scan my portfolio and flag any jobs where cost-to-complete is trending over original budget."* **The Persona AI responds** with a portfolio scan covering all 9 jobs: > **Portfolio health scan, April 17, 2026** > 9 active jobs. \$82.4M total committed value. 3 flagged > > **Harborview Mixed-Use (high risk):** Original budget \$16.4M, committed \$18.2M (+11%). Contingency effectively exhausted at \$42,000. Driven by \$1.1M structural steel overrun plus 3 unsigned change orders (\$184,000). Overrun accelerating from 4% at 50% complete to 11% at 60%. > > **Eastside Warehouse (watch):** 6% over budget at 45% complete. \$480,000 in mechanical extras from late architectural revision. Contingency at 3.6%, runs out at \~65% complete at current pace. > > **Ridgecrest Office (marginal):** 2% over at 72% complete. Minor extras across three trades. Likely closes within contingency. > > **6 other jobs:** All tracking within 1% of original budget. Sarah follows up: *"Show me the change order status for Harborview."* The Persona AI returns the full change order summary showing \$184,000 in unsigned COs, including one that is 18 days overdue (2x the normal turnaround with this owner), and recommends specific escalation actions. **The results:** * 9 jobs health-scored in seconds with trend analysis and contingency burn rate. * \$184,000 in revenue risk surfaced with specific next steps. * Monday morning prep reduced to a 4-minute conversation. ## PM scenario: owner report generation **David Osei - Project Manager** Client report due at 3 pm, normally takes 90 minutes David types: *"Generate an owner progress report for Lakeside Residences. Include budget vs actual, committed costs, schedule status, and top risks."* **The Persona AI generates** a structured report with an executive summary, budget and cost summary, schedule status by phase, pending change orders requiring owner action, and top risks with decision deadlines: > **Lakeside Residences, Owner Progress Report** > > **Executive summary:** Project is 64% complete, tracking within 1.5% of original budget. Framing is 4 days behind baseline due to weather delay. Two items require owner decisions this week. > > **Key figures:** Original contract \$18.4M, revised \$18.55M, projected final \$18.61M (+0.3% variance). Contingency remaining: \$248,000 (1.3%). > > **Top risks:** > > * Window package order deadline Apr 24. 14-week lead time means any delay adds to critical path. > * CO-009 facade changes (\$148,000) due Apr 22. Facade sub cannot mobilize until executed. **The results:** * Full owner progress report generated in under 10 minutes instead of 90. * Data pulled automatically from four Sage modules. ## PE scenario: cost code burn rate alert **Jake Torres - Project Engineer** 2 assigned jobs, 6:45 am site arrival Jake types: *"Which cost codes are burning faster than the schedule percentage on my jobs?"* **The Persona AI responds** with a burn rate analysis across both jobs: > **Cost code burn rate alert, April 17, 2026** > 2 jobs. 47 active cost codes. 5 flagged > > **Harborview Mixed-Use (3 codes at risk):** > > * 03-200 Structural concrete: 72% burned at 60% schedule (+12%). Projected \$148,000 over. > * 09-100 Drywall: 62% burned at 48% schedule (+14%). Labour rate running 8% above estimate. > * 15-400 HVAC ductwork: 67% burned at 55% schedule (+12%). Ductwork rerouting from architectural revision. > > Combined projected overrun: \$305,600 across 3 cost codes. > > **Eastside Warehouse (2 codes to watch):** Both within 5% variance. Flag for weekly monitoring. Jake follows up: *"Draft a change order for the extra excavation work on Eastside."* The Persona AI generates a CO document with actual rates from JCCD cost data, schedule impact assessment, and contract references, ready for PM review. **The results:** * 5 at-risk cost codes identified with root cause analysis. * Change order drafted with actual rates in under 60 seconds. * Morning site prep reduced from 30 minutes to 3 minutes. ## PE scenario: RFI tracking and daily log **Maria Santos - Project Engineer** 3 assigned jobs, end of day on site Maria types: *"Which RFIs are overdue for a response on my jobs?"* **The Persona AI responds** with an overdue RFI tracker: > **Overdue RFI tracker, April 17, 2026** > 3 jobs. 14 open RFIs. 4 overdue > > * **RFI-042** (Harborview, beam depth discrepancy): 10 days overdue. Stop-work risk if not resolved by Apr 21. > * **RFI-038** (Eastside, electrical panel clearance): 5 days overdue. Blocks electrical rough-in. > * **RFI-051** (Northgate, roof drain location): 3 days overdue. > * **RFI-053** (Harborview, fire rating): 2 days overdue. Maria fires off follow-up emails, then types: *"Generate my daily construction log for today."* The Persona AI generates a structured daily log pulling crew data from PMPRJ and PMSUB records, work order completions, equipment on site, weather, and issues, ready for review and submission. ## PM scenario: cash flow forecasting **Elena Vasquez - Project Manager** MEP subcontractor, 5 active jobs, payment timing is critical Elena types: *"Show me my cash flow forecast for the next 30 days across all active jobs."* **The Persona AI responds** with a 30-day forecast: > **30-day cash flow forecast, April 17 to May 17, 2026** > 5 active jobs. Net 30-day position: +\$213,000. Timing risk in week 3 > > Expected inflows: \$847,000 across 5 jobs (confidence scored per client payment history). > Expected outflows: \$634,000 (subs, materials, equipment, insurance). > > **Timing risk, April 24 to May 7:** > A 14-day window where cumulative cash flow goes negative (-\$66,000 peak). > The XYZ Medical payment (\$312,000) is expected Apr 22 but this client > averages 34 days on their last 8 payments. If they pay on their actual > pattern, the gap widens to -\$140,000 for 4 days. > > **Recommendation:** Follow up with XYZ Medical AP today to confirm > payment date. Consider accelerating the Lakeside draw or deferring the > HVAC equipment payment as fallback. **The results:** * \$140,000 cash flow gap identified 10 days before it would have hit. * Payment confidence scored per inflow based on actual client history. * Proactive follow-up with a concrete fallback plan. ## Skills that power all three personas Each persona is powered by a set of skills that define how the Persona AI handles specific types of requests. The Sage Persona AI ships with 24 skills across the three personas, 8 per role. You can trigger a skill in two ways: * Click a **starter prompt** on the home screen to pre-fill the chat input. * Type your request in plain language. The Persona AI auto-detects the appropriate skill. You do not need to memorize skill names or commands. Type what you need in plain language, and the Persona AI matches your intent to the right skill automatically. ### Project Manager skills | **Skill** | **What it does** | | ---------------------- | ------------------------------------------------------------------------------- | | Portfolio Risk Scanner | Scans JCJOB and JCCD records. Ranks jobs by budget variance. | | Profit Forecast | Projects current margin vs at-completion margin per job. | | Owner Report Generator | Assembles progress, cost tables, change orders, and open items into a report. | | Change Order Tracker | Monitors pending change orders by age. Alerts on overdue approvals. | | Schedule Risk Analyser | Calculates float remaining and identifies critical-path delays. | | Cash Flow Forecast | Projects billings vs committed costs over the next 60 days. | | Morning Briefing | Returns top 5 priorities: overruns, stale COs, overdue RFIs, upcoming billings. | | Document Drafter | Generates owner emails, CO follow-ups, and schedule recovery summaries. | ### Project Engineer skills | **Skill** | **What it does** | | ------------------------ | --------------------------------------------------------------------------- | | Job Cost Monitor | Monitors cost codes burning ahead of schedule percentage. | | Budget Overrun Alert | Scans for jobs where actuals exceed budget on assigned jobs. | | RFI Tracker | Tracks open and overdue RFIs. Alerts on missed response deadlines. | | RFI Drafter | Generates structured RFI documents with contract context. | | Change Order Drafter | Creates CO documents from scope and actual JCCD cost data. | | Subcontractor Compliance | Scans for expired insurance and missing lien waivers. | | Daily Log Generator | Generates structured daily logs from crew data and work orders. | | Morning Briefing | Returns top 5 issues: overruns, overdue RFIs, compliance gaps, pending COs. | ### AP Coordinator skills | **Skill** | **What it does** | | ---------------------- | ---------------------------------------------------------------------------- | | Invoice Approval Queue | Pulls pending invoices sorted by aging. | | Over-PO Detector | Flags invoices that exceed their purchase order amount. | | Compliance Checker | Scans for missing W-9s, expired insurance, absent lien waivers. | | Payment Run Builder | Prioritises invoices by due date, discount window, and liability. | | AP Liability Reporter | Aggregates unpaid invoices by job with due-date breakdown. | | Miscoding Detector | Flags job and property codes that do not match active records. | | Vendor Contact Drafter | Drafts W-9 requests, insurance renewal emails, payment confirmations. | | Morning Briefing | Returns top 5 AP priorities: aging invoices, over-PO flags, compliance gaps. | ## Connected to your Sage 300 CRE data The Sage Persona AI connects to your live Sage data via a read-only ODBC service account. No ETL, no nightly sync, no data warehouse required. Each persona only sees the data relevant to their role. An AP Coordinator queries APINV, APVD, JCPO, and PMSUB. A Project Manager queries JCJOB, JCCD, CTJOB, CTCO, and PMPRJ. A Project Engineer queries JCCD, PMRFI, PMSUB, and PMPRJ. Key details: * 10 Sage 300 modules accessed: JCJOB, JCCD, JCPO, CTJOB, CTCO, PMPRJ, PMRFI, PMSUB, APINV, and APVD. * Role-scoped knowledge base per persona. * Dashboards for visual reporting: PM Overview, PE Overview, and AP Overview. * Gmail integration for one-click email drafting. Draft-first, user reviews before sending. * Data queries return in under 4 seconds. Document generation in under 8 seconds. ## Next steps Create and configure role-specific personas in the Admin Console. Assign skills to personas and customize how the Persona AI handles requests. Connect structured data sources and generate knowledge bases for grounded responses. Understand the measurable outcomes the Persona AI delivers. # Welcome to IdeaBoxAI Source: https://docs.ideaboxai.com/index Learn how to make every role in your enterprise AI-powered with persona-driven intelligence, automations, and real-time dashboards. Transform your data into actionable insights with AI-powered analytics and automation. IdeaBoxAI is an intelligent platform that makes every role AI-powered. ## What is IdeaBoxAI? IdeaBoxAI brings together persona-driven AI, real-time dashboards, and workflow automation in a single platform. Whether you need role-specific AI assistants, managed knowledge bases, or automated workflows, IdeaBoxAI gives every function a tailored AI experience. Explore the five core capabilities that power the IdeaBoxAI platform. Role-specific AI assistants with custom personas, skills, and knowledge base integration. AI-powered dashboards and natural language data queries with real-time visualizations. Connect databases, documents, and files with semantic search and source tracing. No-code workflow builder with scheduled and trigger-based execution across departments. ## Getting started Sign up at [app.ideaboxai.com](https://app.ideaboxai.com) to start building dashboards, creating AI assistants, managing data sources, and automating workflows. The following steps walk you through initial setup. Register at [app.ideaboxai.com](https://app.ideaboxai.com) with email or SSO (Google/Microsoft). Tell IdeaBoxAI about your company and functional area, select a persona, and optionally connect your tools โ€” all in under two minutes. Type a natural language prompt and get instant, data-driven insights tailored to your business context. Explore the core capabilities that power the IdeaBoxAI platform. # Creating and querying cubes Source: https://docs.ideaboxai.com/knowledge-bases/cubes Build analytical models with measures, dimensions, and queryable APIs on top of your structured data. Cubes are analytical models built on top of structured data. They define the measures you want to calculate, the dimensions you want to group and filter by, and the SQL logic that generates the data. This guide walks you through creating cubes and executing queries. ## Understanding cubes A cube defines three things. | Component | Description | Example | | -------------- | -------------------------------------------------- | ------------------------------------------- | | **Measures** | Metrics to calculate. | Total revenue, average price, record count. | | **Dimensions** | Attributes to group or filter by. | Date, category, region, status. | | **SQL Logic** | The underlying query that generates the cube data. | A SELECT statement joining relevant tables. | ## Access the cube configuration interface In the structured knowledge base view, the cube interface shows: * A list of existing cubes in the left panel. * The measures and dimensions of the selected cube. * Query execution controls and a result preview. Cube configuration interface showing a list of cubes, measures and dimensions for the selected cube, and query execution controls with result preview. Click **Add Cube** to create a new one. Ensure the underlying datasets are already validated and available before creating cubes. Follow consistent naming conventions to simplify discoverability. ## Create a new cube Add Cube modal showing two creation approaches: AI Generate and Manual Query. The Add Cube modal provides two approaches. ### AI Generate (recommended) Let AI automatically generate the base SQL, measures, and dimensions from a natural language description. Provide a unique, descriptive identifier, for example `ORDER_SUMMARY` or `MONTHLY_REVENUE`. A concise explanation of the cube's analytical purpose. Tell the AI what you want to analyze, for example "Show total revenue by region and product category, with month-over-month trends." Add domain context so the AI can make better decisions about joins, filters, and metric definitions. AI Generate flow for cube creation showing fields for Cube Name, Description, Natural Language Query, and Business Context, with AI-generated SQL and measures below. The AI analyses your datasets and context to generate optimized SQL with appropriate measures and dimensions. ### Manual query For advanced users who want to author SQL directly. Provide a unique identifier for the cube. Author a custom SQL query that defines the cube's data source. Manual Query flow for cube creation showing the Cube Name field and a SQL editor for writing custom queries. Prefer AI Generate for rapid prototyping and exploratory analytics. Use Manual Query for performance-critical or highly customized analytical logic. **Best practices** * Clearly document the business intent in the description and context fields. * Test generated SQL before deploying to production workflows. * Start with simple cubes and iterate based on feedback. ## Execute queries and view results Once a cube is created, you can query it using the cube selection panel. Choose the metrics you want to calculate. Choose how to group and filter the results. Narrow the results to a specific subset of data. Click **Run Query** to execute. Cube selection panel showing measures and dimensions dropdowns, filter controls, Run Query button, and results table. ### View results in multiple formats Query results are available in several formats via tabs. | Tab | Description | | ----------------- | ---------------------------------------------------- | | **Results** | Interactive table or chart view of the query output. | | **Generated SQL** | The SQL query powering the results. | | **SQL API** | An endpoint for SQL-based integrations. | | **REST API** | A RESTful endpoint for application integrations. | | **GraphQL API** | A GraphQL endpoint for data fetching. | **Best practices** * Start with a limited number of dimensions for better performance. * Reuse generated API endpoints instead of re-running UI queries. * Validate results against source systems for critical analytics. * Export and share results with your team. # Knowledge Bases overview Source: https://docs.ideaboxai.com/knowledge-bases/overview Centralize structured and unstructured data for Agentic BI, AI agents, and automation workflows. Knowledge Bases in IdeaBoxAI centralize structured and unstructured data for Agentic BI, AI agents, and automation workflows. This guide covers the complete lifecycle from creation through querying. ## Access the dashboard The Knowledge Bases section is accessible from the **Knowledge Bases** item in the left navigation sidebar. The dashboard displays a searchable table of all existing knowledge bases. The table includes the following columns. | Column | Description | | ----------------- | ----------------------------------------------- | | **Name** | The knowledge base name. Sortable. | | **Status** | Processing state, **Created** or **Processed**. | | **Type** | **Structured Data** or **Unstructured Data**. | | **Created By** | The user who created the knowledge base. | | **Last Modified** | Date and time of the last change. Sortable. | Knowledge Bases dashboard showing a table of knowledge bases with Name, Status, Type, Created By, and Last Modified columns. From the dashboard, you can: * Search for an existing knowledge base by name. * View metadata and processing status at a glance. * Click **+ Create Knowledge Base** to start a new one. Use clear and consistent naming conventions for knowledge bases to help your team quickly identify the right data source for each task. **Best practices** * Review the **Last Modified** column to track stale knowledge bases. * Limit KB creation to meaningful datasets to avoid fragmentation. * Document the purpose and scope of each knowledge base clearly. ## Create a knowledge base ### Select the knowledge base type Click **+ Create Knowledge Base** from the dashboard. A modal presents three types. | Type | Description | Supported formats | | --------------------- | ------------------------------------------------------------------------------------ | --------------------------------------- | | **Unstructured Data** | Data that does not follow a fixed format. Best for semantic search and AI reasoning. | `.pdf`, `.png`, `.jpg`, `.docx`, `.mp4` | | **Structured Data** | Data organized in rows and columns, from SQL databases or spreadsheet files. | `mysql`, `postgres`, `.csv`, `.xlsx` | Knowledge base type selection dropdown showing three options: Unstructured Data, Structured Data, and Financial Documents with supported file formats. The system routes you to a type-specific configuration flow based on your selection. Choose Structured Data only when relational integrity and schemas exist. Use Unstructured Data for documents intended for semantic search or AI reasoning. **Best practices** * Avoid mixing heterogeneous data types in a single knowledge base. * Select the type that matches your data format and intended use case. * Consider how agents will query the data before choosing. ### Configure knowledge base metadata After selecting a type, the **Name & Description** modal appears. Fill in the following fields: | Field | Description | | ------------------- | ---------------------------------------------------------------------------------------- | | **Name** (required) | A unique, descriptive identifier for the knowledge base. | | **Description** | Explains the purpose and scope of the knowledge base. | | **Tags** | Keywords for domain, team, or project association. Type a tag and press enter to add it. | Click **Create** to initialize the knowledge base. Name and Description modal for creating a knowledge base with fields for Name, Description, and Tags. ## Integrating with agents Knowledge bases connect to AI agents. Once connected, agents can: * Query unstructured documents to answer questions. * Run analytical queries on structured data cubes. * Combine multiple knowledge bases for comprehensive insights. * Cite specific sources when providing answers. To connect a knowledge base to an agent, navigate to the agent configuration and select the knowledge base from the dropdown. ## Next steps Upload documents and connect external sources like Google Drive and Confluence. Connect a database, explore schemas, and enrich metadata with AI. Build analytical models with measures, dimensions, and queryable APIs. # Working with structured data Source: https://docs.ideaboxai.com/knowledge-bases/structured-data Connect a database, explore the schema, and enrich metadata with AI for structured knowledge bases. After creating a structured knowledge base, you connect it to a database and explore the imported schema. This guide walks you through entering connection credentials, browsing tables, and enriching metadata. ## Connect your database Enter the connection credentials for your database. | Field | Description | | ----------------- | --------------------------------------------------- | | **Host** | Database server address. | | **Port** | Connection port, for example `5432` for PostgreSQL. | | **Database Name** | The specific database to connect. | | **Username** | Authentication username. | | **Password** | Authentication password. | | **SSL Mode** | Security settings for the connection. | Connect External Data Source dialog showing database type selector and credential fields for Host, Port, Database Name, Username, Password, and SSL Mode. Database credential entry form with fields for Host, Port, Database Name, Username, Password, and SSL Mode, along with connection test and submit buttons. After a successful connection, IdeaBoxAI imports the database schema automatically. ## Explore the Datasets view Once connected, the Datasets view displays a schema visualization of your database. * **Data Sets sidebar**: Lists all imported database tables. * **Table relationships**: Shows joins and foreign key relationships between tables. * **AI enrichment**: Options to enhance metadata with AI, add semantic definitions, and define relationships. Datasets view showing database tables in the sidebar, table relationships in the main panel, and options to enhance with AI, add semantics, and define relationships. From this view you can: * Browse all database tables and their columns. * Define joins and relationships between tables. * Enhance schema metadata using AI assistance. * Document business logic for columns and tables. **Best practices** * Validate relationships against source database constraints. * Avoid circular or ambiguous joins. * Document semantic definitions clearly so agents can reason about your data. The result is a semantically enriched data model ready for analytics and AI queries through cubes. # Working with unstructured data Source: https://docs.ideaboxai.com/knowledge-bases/unstructured-data Upload documents, connect external sources, and monitor processing status for unstructured knowledge bases. After creating an unstructured knowledge base, you need to ingest data from one or more sources. This guide walks you through uploading files, connecting external sources, and monitoring processing status. ## Ingest data from multiple sources After creating the knowledge base, an empty **Data Source** panel appears. Click **Add Data** to see the available ingestion options. | Source | Description | | ---------------------------- | --------------------------------------------------- | | **Upload from Device** | Upload local files such as PDFs and text documents. | | **Add Markdown/Text** | Create or paste text content directly. | | **Upload from Google Drive** | Connect your Google Drive account and import files. | | **Upload from Confluence** | Import pages from your Confluence workspace. | Unstructured data ingestion options showing Upload from Device, Add Markdown/Text, Upload from Google Drive, and Upload from Confluence. From this panel you can: * Upload PDFs and text documents from your local machine. * Connect external sources such as Google Drive and Confluence. * Batch upload multiple files at once. Ensure documents are machine-readable. Avoid scanned PDFs when possible for better processing results. **Best practices** * Upload logically related documents together. * Re-index data after bulk uploads. * Validate file accessibility before uploading from external sources. ## Monitor data processing status After uploading, files enter a processing state. The data source table displays the status of each file. | Column | Description | | ------------- | --------------------------------------- | | **File Name** | The name of the uploaded file. | | **Status** | Current state: Processing or Processed. | | **File Type** | The file format: PDF, TXT, and others. | From this view you can: * Monitor ingestion and processing status in real time. * Trigger manual re-indexing if required. * Remove failed files and re-upload them. Wait for all files to reach **Processed** status before querying. Investigate repeated processing failures promptly. Once all files are processed, the data becomes searchable and usable by agents, Agentic BI, and automation workflows. # Enterprise integrations Source: https://docs.ideaboxai.com/partnerships/ibm/enterprise-integrations Learn which enterprise tools the IdeaBoxAI Sales Persona AI connects to and how the IBM integration framework works. Sales Persona AI is not just a standalone AI system. By integrating IBM's core connectivity technologies, it is deeply embedded within your existing enterprise ecosystem. This seamless integration ensures that the AI agent can access, update, and synchronise critical business data in real time. ## Core integration platform and use cases Sales Persona AI works best when users understand which connected system should answer which part of a sales question. In watsonx Orchestrate, the IdeaBoxAI Sales Persona AI agent brings the AE, SE, and SDR subagents together with the approved business tools below. Users do not need to choose a tool manually in most conversations. They can describe the business outcome, and Persona AI will select the appropriate available tool based on the request and required input fields. ### CRM and task management (Salesforce and Jira) Salesforce is the system of record for account, case, event, and opportunity-related context. Jira is the operational source for implementation work, product issues, technical blockers, and POC execution details. **Typical first-use scenarios:** * **AE preparation**: Ask Persona AI to retrieve recent Salesforce cases for an account before a renewal, expansion, or executive alignment meeting. * **SE risk review**: Ask Persona AI to inspect Jira issues or projects connected to a POC so unresolved technical blockers can be summarised before the next customer call. * **SDR prioritisation**: Ask Persona AI to check CRM activity or recent events before drafting a targeted outreach message. * **Deal hygiene**: Ask Persona AI to compare open customer concerns against known Jira work so the sales team can identify what still needs an owner. **Example prompt:** > *"I have a renewal call with Acme Corp. Review recent Salesforce cases and summarise the top risks I should address as the AE."* ### Omnichannel communication and collaboration (Gmail, Outlook, and Teams) Sales communication rarely lives in one inbox. Sales Persona AI lets users send email through Gmail or Outlook and review messages across available mail folders. Users can also inspect Microsoft Teams calendar and team information from the same conversational workspace. **Typical first-use scenarios:** * **Meeting follow-up**: Ask Persona AI to draft and send a concise follow-up email after a discovery call, using the action items already discussed in the conversation. * **Inbox triage**: Ask Persona AI to review recent customer replies in Outlook and summarise objections, requested next steps, or buying signals. * **Meeting preparation**: Ask Persona AI to check upcoming Teams events and prepare a role-specific briefing for the AE, SE, or SDR. * **Stakeholder alignment**: Ask Persona AI to turn CRM or Jira context into a customer-ready message and send it through the approved email channel. **Example prompt:** > *"Draft a Gmail follow-up for the customer after today's POC review. Keep it under 150 words, include the three agreed action items, and ask for confirmation on the Q3 timeline."* ### Prospective customer background checks and data enrichment (ZoomInfo) Sales Persona AI uses ZoomInfo to enrich prospect and account profiles with company, contact, and trigger-event context. **Typical first-use scenarios:** * **Prospect research**: Ask Persona AI to look up a target company and summarise why it may be a good fit. * **Decision-maker discovery**: Ask Persona AI to identify likely stakeholders for an outbound sequence or account plan. * **Trigger-based outreach**: Ask Persona AI to use recent business signals to personalise an email or LinkedIn-style message. * **ICP scoring**: Ask Persona AI to evaluate whether a target account appears to match the ideal customer profile before spending time on outreach. **Example prompt:** > *"Research Acme Corp with ZoomInfo, identify likely decision-makers, and give me three personalised SDR outreach angles."* ## Technical advantages: powered by the IBM integration framework The IBM integration framework provides several enterprise-grade advantages for Sales Persona AI deployments. * **Secure by default**: Connections to third-party systems are handled through the IBM integration layer, using enterprise authentication and encrypted transport. Users can focus on the sales task while administrators retain control over authorisation, credential management, and tenant boundaries. * **Schema-guided tool execution**: Each connected tool in watsonx Orchestrate exposes a clear input schema. Sales Persona AI uses that schema to determine what information is required and what is optional. If additional details are needed, then it asks the user a follow-up question before taking action. * **Role-aware context handling**: The same source data can produce different outputs depending on whether the user is acting as an AE, SE, or SDR. For example, a Salesforce case history can become an AE renewal risk summary, an SE technical objection list, or an SDR account research note. * **Reduced context switching**: Users can retrieve data, summarise findings, draft messages, and trigger approved communication actions from one watsonx Orchestrate conversation. There is no need to move manually across Salesforce, Jira, Gmail, Outlook, Teams, and ZoomInfo. * **Traceable business actions**: Read actions, such as retrieving Salesforce cases or listing Outlook messages, are separated from write actions, such as sending an email. New users can start with low-risk research tasks before asking Persona AI to perform outbound communication. * **Composable workflows**: Users can chain actions naturally. A single session can start with Salesforce case review, continue into Jira blocker analysis, and end with a Gmail or Outlook follow-up draft. Persona AI preserves the conversation context needed to avoid repeated manual input. # Maintenance Source: https://docs.ideaboxai.com/partnerships/ibm/maintenance Current system status and scheduled maintenance information for Sales Persona AI. ## System Status All services are operating normally. ## Scheduled Maintenance Notice To ensure the continued optimization of Sales Persona AI's multi-agent collaboration performance and security, we perform regular system upgrades and maintenance. * **Maintenance Frequency:** We conduct maintenance on an irregular basis in accordance with the product update cycle, typically during off-peak business hours. * **Advance Notice:** For any scheduled maintenance that may affect service availability, we will notify affected users or administrators via official channels at least 24โ€“48 hours in advance. # Overview Source: https://docs.ideaboxai.com/partnerships/ibm/overview Learn how the IdeaBoxAI Sales Persona AI brings role-based sales intelligence to IBM watsonx Orchestrate with enterprise-grade integrations. Sales Persona AI is an enterprise-grade multi-agent AI sales enablement system developed by IdeaBoxAI. More than just a tool, it serves as a digital sales think tank equipped with deep business logic. It integrates three highly specialized AI roles: Account Executive (AE), Sales Engineer (SE), and Sales Development Representative (SDR). Together, these roles provide automated and intelligent support across the entire sales lifecycle, from lead generation to deal closure. Sales Persona AI delivers three core benefits to the sales team. Deeply integrates with every key touchpoint in the sales cycle, including pre-meeting research, technical discovery, POC planning, objection handling, and forecast management. All outputs such as emails, briefings, and talking points are strictly based on real business signals from sources like Salesforce. No false or generic content is generated. Automatically syncs research findings, meeting minutes, and email drafts to CRM and cloud storage, significantly reducing the manual data entry burden on sales representatives. ## Why choose Sales Persona AI? Sales Persona AI represents a generational leap in research efficiency, reducing preparation time from hours to minutes. It provides precise technology fit assessments and dynamic objection tracking. These capabilities help sales teams overcome conversion bottlenecks and drive measurable sales results. ## Next steps Explore the following guides to learn how the IBM integration works and get started with Sales Persona AI. See which tools Sales Persona AI connects to and how the IBM integration framework works. Subscribe, connect your accounts, and start using Sales Persona AI in watsonx Orchestrate. # Quick start guide Source: https://docs.ideaboxai.com/partnerships/ibm/quick-start-guide Subscribe to the IdeaBoxAI Sales Persona AI on IBM Cloud, connect your accounts, and start your first conversation. Follow these steps to subscribe, connect your accounts, and start using Sales Persona AI in IBM watsonx Orchestrate. ## Subscribe To use Sales Persona AI in your enterprise environment, you need to subscribe to the service and create resources through the IBM Cloud Catalog. This process ensures that your account is licensed for a compliant service instance. Log in to your IBM Cloud console. Click **Catalog** in the top navigation bar. Enter **Sales Persona AI** in the search bar. Click the Sales Persona AI tile to go to the details page. * **Select a plan**: From the **Pricing** list, choose a plan that suits your business size (for example, a usage-based plan). * **Set instance information**: In the right-hand panel, enter the **Service name**, select the **Resource group**, and choose the deployment region. Review the cost estimate on the right side of the page. Read and agree to the Third-Party Terms of Service and the EULA. Click **Create**. When you click **Create**, IBM Cloud immediately sends an API call to the IdeaBoxAI Service Broker. The system automatically assigns you a unique Resource ID (CRN) and API key. It then provisions the corresponding multi-agent computing resources in the IdeaBoxAI backend. Once the instance is provisioned, find it in the **Resource List** on IBM Cloud. Click the instance name to go to the **Manage** page and view the control panel. ## Connection and authorisation Although the agent is ready, you must complete the OAuth authorisation for your personal account in **Connections** so that it can process your Salesforce records or send Gmail. Go to **Manage** then **Connections** in the watsonx Orchestrate sidebar. IBM watsonx Orchestrate sidebar showing the Manage menu expanded with Connections, Voice, Phone, and Access management options. Locate the connections related to the Sales Persona AI tools such as Salesforce, Jira, Google, and Microsoft. In the official login window that appears, enter your business account credentials and grant authorisation. Once successful, the connection status must display as **Live**. Only then can the agent perform read and write operations on your behalf. You may refer to [Connecting apps for prebuilt agents - IBM Documentation](https://www.ibm.com/docs/en/watsonx/watson-orchestrate/current?topic=apps-connecting-prebuilt-agents) for more details. In the chat box, try asking: *"Please list my top 3 opportunities in Salesforce this week, AE Assistant."* If the system returns the data accurately, then authorisation is complete. ## Using the tools Each tool registered on watsonx Orchestrate exposes a defined input schema. When you talk to Sales Persona AI, the agent requests the minimum fields required by that schema. You can either provide every parameter in a single instruction or let the agent ask targeted follow-up questions when something is missing. For the best first experience, start with a read action before using a write action: 1. Ask Persona AI to retrieve or summarise information from Salesforce, Jira, Outlook, Teams, or ZoomInfo. 2. Review the result and correct any account name, date range, stakeholder, or business context if needed. 3. Ask Persona AI to turn that context into a customer-facing email, internal update, or next-step checklist. 4. Only then ask Persona AI to send the message through Gmail or Outlook if the content is ready. This flow helps new users build confidence because they can see exactly what information Persona AI found before asking it to take an external action. ### How to ask Persona AI effectively Use plain business language, but include the fields a sales operator would normally need. | Field | What to include | | ----------- | ---------------------------------------------------------------------------------------------------------------- | | **Who** | Account name, customer contact, internal owner, or recipient email address. | | **What** | The business outcome, such as "prepare a renewal briefing", "summarise open cases", or "draft a POC follow-up". | | **When** | Date range, upcoming meeting date, recent messages, or priority deadline. | | **Channel** | Salesforce, Jira, Gmail, Outlook, Teams, or ZoomInfo if you already know where the information should come from. | | **Role** | AE, SE, or SDR when you want the answer framed for a specific sales responsibility. | ### Example 1: Sending an email via Gmail (a write action) The **Send an email using Gmail** tool requires three required inputs: the recipient address (`to`), the subject line, and the message body. Send an email using Gmail tool parameters showing required fields for email_address, subject, and body, and optional fields for cc_email_address and bcc_email_address. A natural, single-turn instruction covers all three: > *"Compose and send an email from my Gmail account to [jane.doe@client.com](mailto:jane.doe@client.com). The subject line should be Follow-up on Yesterday's POC Review, and the email body should read: Please finalise the data solution as soon as possible."* If you omit one of the required fields, then Persona AI will ask a targeted follow-up question to fill the gap. For example, if you forget the subject, then it will prompt you for one. Optional fields such as `cc` and `bcc` can be included by mentioning them in the instruction (for example, "also cc the SE lead at [alex@client.com](mailto:alex@client.com)"). ### Example 2: Retrieving Salesforce cases (a read action) The **Get list of cases from Salesforce** tool accepts an optional search query and pagination controls (`limit`, `offset`). Get list of cases from Salesforce tool parameters showing fields for current_run, limit, search, and skip. You can start broad and narrow down: > *"List the 10 most recent open Salesforce cases that mention 'Acme Corp' so I can prepare for my renewal call tomorrow."* Persona AI will invoke the tool, apply the search filter, and return a structured summary. You can then chain follow-up instructions without repeating any context. For example, ask Persona AI to draft a Gmail message to the account owner summarising the open cases. ### Example 3: Preparing for a technical POC review with Jira The Jira tools help SE users understand implementation status and technical risk before customer-facing meetings. Start with a focused question: > *"Review the Jira issues for the Acme POC project and summarise unresolved technical blockers, owners, and recommended next steps for the SE."* Persona AI will use the available Jira project or issue information to produce a concise risk summary. If the project name or filter is ambiguous, then it will ask a follow-up question before retrieving the data. ### Example 4: Researching a prospect with ZoomInfo ZoomInfo is most useful when the user needs account enrichment before outreach or account planning: > *"Use ZoomInfo to research Acme Corp. Summarise the company profile, likely decision-makers, and three SDR outreach angles based on recent business signals."* After Persona AI returns the research summary, you can continue with: > *"Turn the strongest outreach angle into a short first-touch email for the VP of Sales."* ### Example 5: Preparing from Teams and Outlook context When meeting and message context are spread across collaboration tools, ask Persona AI to collect the latest signals first: > *"Check my upcoming Teams meetings for Acme Corp this week and review recent Outlook messages from that account. Prepare an AE briefing with risks, open questions, and recommended talking points."* If Persona AI needs a folder name, date range, or contact address to complete the request, then it will ask for that missing field instead of guessing. ### Recommended first-session workflow New users should try the following sequence in watsonx Orchestrate: 1. **Start with account context**: *"Summarise what I should know about Acme Corp before my next meeting."* 2. **Add system-specific evidence**: *"Include recent Salesforce cases and any Jira blockers related to the POC."* 3. **Choose a role lens**: *"Frame the summary for an AE"* or *"Frame the summary for an SE."* 4. **Create an output**: *"Draft a customer follow-up email with the next steps."* 5. **Review before sending**: Read the generated message, then ask Persona AI to send it through Gmail or Outlook only when it is ready. ### Working with missing information If a request is incomplete, Persona AI will ask for only the missing fields required by the tool. For example: * If an email request lacks a recipient, then Persona AI asks for the recipient address. * If a Salesforce query is too broad, then Persona AI may ask for an account name, case status, or result limit. * If a Jira request does not identify a project or issue, then Persona AI may ask which project should be searched. * If a Teams or Outlook request needs a time window, then Persona AI may ask for the date range. This guided behaviour is intentional. It keeps first-time users from needing to memorise tool schemas while still ensuring that tool calls remain accurate. ### Discovering the input schema for any tool Every tool listed in the agent's Toolset follows the same pattern. This includes Outlook send/read, Microsoft Teams, Jira issues and projects, Salesforce cases and events, Gmail send, and ZoomInfo enrichment. Open the tool entry on watsonx Orchestrate to view its required and optional input fields. Then phrase your instruction to Persona AI so those fields are either stated explicitly or inferred from prior conversation context. When in doubt, describe the business outcome instead of the technical operation. For example, say *"prepare me for the renewal call"* rather than *"call the Salesforce case API."* Persona AI will map the request to the available tool, ask for any missing input, and return the result in the sales role format you need. ## Starting a conversation The core interaction logic of Sales Persona AI is based on natural language. You do not need to learn complex code or fixed commands. Simply input your business requirements into the watsonx Orchestrate chat window as if you were communicating with a professional sales assistant. ### Persona-based triggers The system automatically identifies and invokes the most appropriate role-based assistant based on your instructions. For the most precise response, it is recommended to explicitly reference the role in your request. **Collaborating with the AE Assistant.** Focused on deal strategy, meeting preparation, and risk assessment. > *"Ask the AE Assistant to prepare a briefing for my meeting with Acme Corp this afternoon and analyse potential deal risks."* **Collaborating with the SE Assistant.** Focused on technical architecture, rebuttal handling, and POC planning. > *"Regarding the 'data scalability' doubt from the client, ask the SE Assistant to generate an evidence-backed rebuttal card."* **Collaborating with the SDR Expert.** Focused on lead generation, personalised outreach, and data enrichment. > *"Ask the SDR Expert to draft 3 personalised outreach angles based on the latest funding signals from ZoomInfo."* ### Multi-agent workflows Due to the system's context-awareness, you can switch roles seamlessly within the same conversation window without repeating client background information. > *After the AE summarises customer pain points, you can follow up directly: "Based on these pain points, ask the SE Assistant to generate a tailored technical discovery questionnaire."* ### Guided interactions If your instruction lacks sufficient information to invoke a specific tool, then Persona AI will proactively ask follow-up questions to fill the gaps. This applies to actions such as sending an email or querying a specific record. * If you request to send an email without providing a recipient, it will ask: *"To which contact's email address should this message be sent?"* * If your query is too broad, it may suggest: *"Should I retrieve the 10 most recent or all related Salesforce cases?"* ### First-session tips To help you quickly build trust and familiarise yourself with the process, follow the **read-before-write** principle during your first session. 1. **Research**: First, ask Persona AI to retrieve and summarise existing data. For example: *"Summarise recent POC blockers recorded in Jira."* 2. **Content generation**: Once you have confirmed the research is accurate, ask it to generate an output. For example: *"Turn these blockers into a follow-up email draft for the client."* 3. **Final execution**: Issue the send command only when you are satisfied with the draft. # Support Source: https://docs.ideaboxai.com/partnerships/ibm/support Get help with the Sales Persona AI for IBM watsonx Orchestrate. For questions or issues with the Sales Persona AI for IBM watsonx Orchestrate, contact [support@ideaboxai.com](mailto:support@ideaboxai.com). # Overview Source: https://docs.ideaboxai.com/partnerships/overview Learn how IdeaBoxAI partners with leading enterprise platforms to deliver AI-powered sales intelligence and automation. IdeaBoxAI works with industry-leading platforms to bring AI-powered intelligence directly into the tools your team already uses. Each partnership delivers a purpose-built integration that connects IdeaBoxAI capabilities to a specific enterprise ecosystem. Explore the available partnerships below. IBM logo Sales Persona AI for IBM watsonx Orchestrate brings role-based sales intelligence to your enterprise ecosystem. It connects to Salesforce, Jira, Gmail, Outlook, Microsoft Teams, and ZoomInfo through IBM's secure integration framework. # Platform overview Source: https://docs.ideaboxai.com/platform-overview Learn how IdeaBoxAI's core capabilities work together to make every role in your enterprise AI-powered. IdeaBoxAI brings together persona-driven AI, real-time dashboards, and agentic BI in a single platform. Each capability is designed to give every function in your enterprise a tailored AI experience. ## Platform capabilities The following capabilities form the foundation of the IdeaBoxAI platform. Build role-specific AI for every function in your enterprise. Assign role-specific skills instead of generic prompts and deploy them to your team, ready to run. Build and deploy role-specific AI skills, the core building blocks that make each persona capable. Use pre-built skill templates or create custom skills with no code required. Ask questions in plain English and get live, SQL-backed answers from your own systems. No analyst queue, no stale reports. Every persona gets real-time visibility into what matters to their role. Let AI agents explore your data, surface insights, and answer follow-up questions on their own. Every persona gets a self-driving analyst that turns business questions into live, SQL-backed answers. Ground every AI in your institutional knowledge: policies, documents, and data. Semantic search across all content with every answer traced back to its source. # Quickstart Source: https://docs.ideaboxai.com/quickstart Learn how to create your IdeaBoxAI account, complete onboarding, and launch your workspace. This guide walks you through registering for an IdeaBoxAI account, completing the three-step onboarding wizard, and launching your workspace. By the end, you will have a fully configured workspace with personas and connected tools. ## Create your account To get started, navigate to the [IdeaBoxAI login page](https://app.ideaboxai.com/auth/login). The registration screen gives you two ways to create an account: email and password or single sign-on. IdeaBoxAI login and registration screen showing email and password fields, sign in button, and Google and Microsoft SSO options. ## Register with email Follow these steps to create an account using your email address. Go to [app.ideaboxai.com/auth/login](https://app.ideaboxai.com/auth/login) and click the **Sign up** link at the bottom of the form. The Create your account page appears. IdeaBoxAI sign-up page showing Username (Email), First Name, Last Name, Password, and Confirm Password fields, Sign Up button, and Google and Microsoft SSO options. Fill in the following fields: * **Username (Email)**: Your email address. * **First Name** and **Last Name**: Your full name. * **Password**: Choose a password. * **Confirm Password**: Re-enter your password. Then click **Sign Up** to create your account. Check your inbox for a verification email from IdeaBoxAI. Click the link to confirm your account. ## Register with SSO If your organisation uses Google or Microsoft, you can register with a single click. Click the **Google** or **Microsoft** icon below the sign-in button to authenticate with your existing account. No separate password is required. ## Sign in to an existing account If you already have an account, enter your email and password on the [login page](https://app.ideaboxai.com/auth/login) and click **Sign In**. You can also sign in using **Google** or **Microsoft** SSO by clicking the relevant icon below the sign-in button. If you forget your password, click **Forgot Password?** to receive a reset link by email. ## Onboarding wizard After signing in for the first time, IdeaBoxAI walks you through a three-step setup wizard. The left sidebar tracks your progress across each step. Want to get started immediately? Click the **Quick Launch** button at any point during onboarding to skip the remaining steps and jump straight into your workspace with default settings. You can always update your profile, personas, and connections later from Settings. ### Step 1: About You This step takes about 30 seconds. IdeaBoxAI collects basic information about you and your business to tailor the workspace experience. About You onboarding step showing Company Name, functional area, website URL, and AI-generated business summary fields. Fill in the following fields: * **Company Name**: Your organisation's name. * **Functional area**: Select the department or function you work in (e.g., HR, Sales, Operations). This determines which personas and skills IdeaBoxAI recommends. * **Website URL**: Enter your company website. IdeaBoxAI uses this to auto-generate a business context summary. Once you enter the website URL, IdeaBoxAI analyses it and generates an **AI business summary** describing what your company does. You can review and edit this summary before continuing. About You step with AI-generated business summary populated from the company website. Check the **Terms & Conditions and Privacy Policy** checkbox, then click **Continue** to proceed. You can also click **Quick Launch** to skip ahead with defaults. ### Step 2: Your Workspace This step takes about 1 minute. IdeaBoxAI presents your AI Business Context and recommends personas based on the functional area you selected. Your Workspace step showing AI Business Context card and recommended personas for the selected functional area. At the top, the **AI Business Context** card shows the profile IdeaBoxAI built from your inputs. Click **Edit** to refine it or the refresh icon to regenerate. Below that, you see persona recommendations tailored to your functional area. Each persona comes with pre-configured skills: * **Select a persona** by clicking the radio button next to it. You must select at least one to continue. * **View abilities** by expanding the Abilities dropdown on any persona to see the skills it includes. * **Create Your Own** if none of the suggestions fit. Click this option to build a custom persona by answering a few questions. Click **Continue** to proceed or **Back** to return to Step 1. ### Step 3: Connect & Go This final step lets you choose how to populate your workspace with data. Connect and Go step showing two options: Continue with sample data and Connect your tools. You have two options: **Continue with sample data** โ€” Jump straight into your workspace with realistic, pre-loaded data for your functional area. This lets you explore every feature instantly with no setup. You can connect real tools later from Settings. * Pre-loaded dashboards and knowledge bases * Try every feature instantly * Connect real tools later **Connect your tools** โ€” Bring in your real CRM, email, and data sources for personalized insights from day one. This takes 2-3 minutes and is the **recommended** path. * Use live data from your stack * Personalized recommendations * Takes 2-3 minutes #### Connecting your tools If you choose **Connect your tools**, IdeaBoxAI shows a tool selection screen with categories recommended for your functional area. Tool connection screen showing categories like Email, Calendar, and Documents with recommended tools for the selected functional area. * **Search** for a specific tool using the search bar at the top. * **Filter by category** by selecting from the available category tabs. The categories shown vary based on your functional area and role. * **Recommended tools** appear at the top, tailored to your functional area (e.g., Gmail, Google Calendar, and Google Drive for an HR team). * Click **Show more categories** to see additional tool categories. Select the tools you want to connect, then click **Launch workspace** to finish setup. You can always add or remove tools later from **Settings > Connections**. ## Your workspace After completing the onboarding wizard, you land in your IdeaBoxAI workspace. A guided tutorial walks you through the key features in 6 steps. IdeaBoxAI workspace home screen showing the welcome tutorial popup and the co-pilot chat interface. The workspace includes: * **Persona AI chat**: Ask anything using natural language. Skills are auto-detected, or type `/` to pick one. * **Skill suggestion cards**: Quick-action cards based on your persona and functional area. * **Sidebar navigation**: Access the Assistant, Dashboards, Personas, Skill Studio, Knowledge Bases, Connections, and Profile pages. Your profile is automatically configured with the persona you selected during onboarding. You can view and update your display name, personas, and language from **Settings > Profile**. Profile settings page showing avatar, display name, assigned personas, language, and last active timestamp. ## Next steps Explore the five core capabilities that power IdeaBoxAI. Learn how the Persona AI delivers role-specific intelligence for your team. # Frequently asked questions Source: https://docs.ideaboxai.com/resources/faq Comprehensive FAQ covering product overview, security, architecture, integrations, deployment, pricing, and competitive differentiation. This reference covers the most common questions about IdeaBoxAI, organized by topic. Whether you are a first-time visitor, a security reviewer, a solution architect, or a procurement lead, you can jump to the section most relevant to you. ## Product overview IdeaBoxAI is an enterprise AI platform that helps organizations deploy purpose-built AI assistants, called Personas, for specific business roles and workflows. Rather than providing one general-purpose chatbot, IdeaBoxAI lets companies configure AI assistants that understand a defined job, follow company-specific policies, and connect securely to internal systems. Consumer AI tools are designed for broad, general use and have no awareness of your company's data, policies, or workflows. IdeaBoxAI is built for enterprise deployment: each Persona is grounded in your organization's documents, governed by configurable rules, and integrated with your internal systems. It is intended as a business platform rather than a consumer assistant. End users do not need coding skills. Interacting with a Persona is similar to messaging a colleague. Administrators and builders who configure Personas benefit from familiarity with business processes and basic data concepts, but a low-code interface is provided for most configuration tasks. No. While Personas can be used through a chat interface, the platform also supports task execution, document retrieval, integrations with business systems, scheduled workflows, and API-driven automation. Many Personas operate without any chat interaction at all. IdeaBoxAI is designed to augment employees by handling repetitive, time-consuming tasks. Customers typically use it to reduce administrative overhead, accelerate research, and free staff to focus on higher-value work. As with any productivity technology, individual organizations decide how to apply it within their workforce strategy. A Persona is a configured AI assistant with a defined role, scope of knowledge, set of permitted actions, and behavioral rules. Instead of a single generic AI, your organization may operate several Personas โ€” for example, an HR Assistant, a Sales Enablement Assistant, or an IT Support Assistant โ€” each tuned to its specific function. Yes. Users can type or speak to a Persona using everyday language. The platform also supports structured inputs, forms, and API calls for automated scenarios. IdeaBoxAI is built for mid-market and enterprise organizations that want to deploy AI in a controlled, auditable way. Typical customers include companies in financial services, healthcare, professional services, manufacturing, and the public sector. Customers commonly report faster handling of routine tasks, fewer errors in repetitive processes, more consistent application of internal policies, and better self-service access to institutional knowledge. Actual results depend on the use case and the quality of the underlying data. The platform is web-based and works in modern browsers on desktop and mobile devices. Specific access methods depend on how your organization configures authentication and network policies. IdeaBoxAI is a United Statesโ€“based company. Customer-facing operations, engineering, and support are organized to serve a global customer base, with regional data residency options available for customers with applicable requirements. Many organizations want to adopt AI but struggle with three issues: keeping sensitive data private, ensuring AI behavior aligns with internal policy, and grounding AI responses in their own information rather than the public internet. IdeaBoxAI provides the controls, retrieval architecture, and governance layer needed to address these concerns. No. IdeaBoxAI does not train or sell foundation models. It is a platform that orchestrates language models from providers such as OpenAI, Anthropic, Google, and open-source options, and adds the enterprise capabilities required to use them responsibly. ## How IdeaBoxAI works A Persona is a configured AI assistant defined by four elements: its role description, the knowledge sources it may access, the tools or actions it is permitted to use, and the operational rules it must follow. Personas are the central unit of configuration in IdeaBoxAI. Yes. Administrators and approved builders can create Personas tailored to specific departments, teams, or workflows. The platform provides templates for common roles to accelerate setup. Yes. A Persona can be configured to delegate sub-tasks to another Persona โ€” for example, a research-focused Persona can pass findings to a drafting-focused Persona. All inter-Persona communication respects each Persona's access controls. It declines politely and, where appropriate, suggests an alternative Persona or escalation path. Administrators can configure how the Persona responds to out-of-scope requests. Yes. Personas, along with their memory and configuration, can be archived or permanently deleted by administrators. Audit records of past activity are retained according to your organization's retention policy. Long-term memory refers to the persistent knowledge store that allows a Persona to recall prior interactions, reference documents, and project context across sessions. Memory is implemented using retrieval-augmented generation (RAG) backed by a vector database and a metadata layer. Personas do not learn in the same sense as a model being retrained. However, administrators can update guidelines, add or correct knowledge sources, and capture feedback that the Persona uses on subsequent interactions. When a query is received, the Persona retrieves the most relevant content from its configured knowledge sources based on semantic similarity and metadata filters, then uses that content as grounding for the response. Yes. Administrators control whether a Persona has access to long-term memory, what data is included, how long it is retained, and whether memory is shared across users or kept private to each user. Documents can be tagged with effective and expiration dates. The retrieval layer can be configured to deprioritize or exclude stale content, and bulk re-indexing tools are provided for periodic refreshes. Skills are the actions a Persona is permitted to take beyond responding in natural language โ€” for example, querying a CRM, creating a ticket, sending an email, or running an internal API. Each Skill is defined with explicit inputs, outputs, and permissions. Operational rules define the boundaries of acceptable behavior โ€” what topics are in scope, which Skills may be used under which conditions, and how to handle sensitive scenarios. They function as a configurable policy layer that the Persona must follow. The platform's governance layer evaluates each request against the Persona's rules. Where a conflict exists, the rule takes precedence and the Persona explains, at an appropriate level of detail, why the request cannot be fulfilled. Yes. Rules are layered: organization-wide policies, role-specific guidelines, and conversation-level instructions can all apply simultaneously. The platform documents which rules influenced any given response for auditability. ## Comparisons with other AI tools ChatGPT Enterprise is a productivity assistant centered on a chat experience powered by OpenAI models. IdeaBoxAI is an orchestration and governance platform that operates above the model layer, supports multiple model providers, and is structured around role-specific Personas with explicit policy controls and integrations. Custom GPTs allow users to package instructions, files, and limited actions for a specific use case within ChatGPT. IdeaBoxAI Personas include the same kinds of capabilities and add features commonly required in enterprise settings: granular role-based access control, configurable retention, multi-model routing, audit logging, and structured workflow execution. Yes. OpenAI models are one of several supported providers. Administrators choose which model or models a given Persona may use, subject to your organization's contractual relationships with each provider. Some customers offer ChatGPT Enterprise for general productivity while using IdeaBoxAI for role-specific, system-integrated workflows that require stricter governance and connections to internal data. Claude for Work, from Anthropic, provides access to Claude models with team features and Projects for grouping context. IdeaBoxAI is a multi-model platform that can use Claude as one of its backends while adding orchestration, system integrations, role-based access control, and policy enforcement beyond a single-vendor chat product. Yes. Claude is among the supported model families. Customers can route specific Personas, or specific task types within a Persona, to Claude based on cost, latency, or task-fit considerations. Claude Projects let users group documents and instructions for a particular topic. IdeaBoxAI provides a broader retrieval architecture โ€” including hybrid search, metadata filtering, document-level access controls, and re-ranking โ€” designed for larger and more sensitive corpora than a single Project context. Microsoft 365 Copilot is integrated tightly with Microsoft 365 applications and the Microsoft Graph. IdeaBoxAI is application-agnostic and integrates with a wide range of business systems, including Microsoft 365, Google Workspace, Salesforce, ServiceNow, and others, through configurable connectors. Copilot Studio is a builder for conversational agents within the Microsoft ecosystem. IdeaBoxAI is a standalone platform that runs on the customer's preferred cloud or on-premises infrastructure and supports a broader set of model providers, integrations, and deployment patterns. Yes. Connectors to Microsoft 365 services โ€” including SharePoint, OneDrive, Outlook, and Teams โ€” are available, subject to the customer's licensing and tenant configuration. IdeaBoxAI sits above the foundation model layer. Its value comes from orchestration, retrieval, governance, integrations, and lifecycle management โ€” capabilities that complement rather than replace the underlying models. No. The platform is designed to be model-agnostic. Customers can switch or mix providers without re-engineering Personas or workflows, subject to each provider's terms. Hallucinations are reduced โ€” though not eliminated โ€” through retrieval grounding, citation of source documents, and configurable strict-grounding modes that instruct the model to respond "insufficient information" when retrieved sources do not support an answer. Outputs should still be reviewed for accuracy in high-stakes contexts. ## Security, privacy, and governance It means security, access control, and policy enforcement are built into the platform's request lifecycle rather than added as an afterthought. Every request passes through authentication, authorization, policy evaluation, and logging layers before reaching a language model. Customer data is not used to train foundation models. IdeaBoxAI configures its provider integrations to opt out of model training where the provider supports it, and recommends that customers verify these settings as part of their own due diligence. Data is encrypted in transit using TLS 1.2 or higher and at rest using AES-256 or equivalent. Customer-managed encryption keys are supported on applicable deployment options. Storage location depends on the chosen deployment model. Multi-tenant cloud deployments offer region selection; private cloud deployments use the customer's chosen region; on-premises deployments keep data within the customer's environment. Yes. IdeaBoxAI supports multi-tenant SaaS, single-tenant private cloud (in the customer's AWS, Azure, or Google Cloud account), and on-premises installations for customers with strict data residency or air-gapped requirements. The platform supports SAML 2.0 and OpenID Connect (OIDC) for single sign-on with providers such as Microsoft Entra ID, Okta, Ping, and Google Workspace. SCIM is supported for automated user provisioning and deprovisioning. Yes. Access to Personas, knowledge sources, Skills, and administrative functions is controlled through configurable roles. Attribute-based controls โ€” for example, restricting a Persona to users in a specific department or region โ€” are also supported. Administrative actions require elevated permissions and are logged in an immutable audit trail. Customers can require multi-factor authentication for all administrative sign-ins and can integrate with privileged access management tools. The platform is designed to support customers' GDPR obligations through features such as data residency selection, configurable retention, data subject access workflows, and a Data Processing Addendum. Compliance with GDPR remains a shared responsibility between IdeaBoxAI and the customer. IdeaBoxAI offers configurations suitable for use with protected health information for customers who require them, including the execution of a Business Associate Agreement on qualifying deployments. Customers should confirm the latest scope and applicable deployment options with their account team. IdeaBoxAI maintains a security program aligned with established frameworks. The current list of attestations, audit reports, and certifications is available through the Trust Center and updated as new audits are completed. Administrators can configure detection and masking of common PII categories in prompts, retrieved content, and stored logs. Detection policies are configurable; perfect identification of all PII in free text cannot be guaranteed. Defenses include input validation, separation of system instructions from user content, output filtering against policy, and tool-use authorization checks. No single control fully eliminates prompt injection; IdeaBoxAI follows current industry guidance and updates defenses as the threat landscape evolves. Audit logs record authentication events, configuration changes, Persona usage, Skill invocations, and policy decisions. Logs can be exported to customer-managed SIEM systems and are retained according to configurable retention policies. Administrators with appropriate permissions can review usage data for governance and quality purposes. Privacy controls โ€” including configurable redaction of message contents in admin views โ€” allow customers to balance oversight with employee privacy obligations. Yes. Output policies can be configured to refuse requests for system instructions, and review of suspicious outputs is supported. As with prompt injection, no defense is absolute, and operators should treat sensitive system prompts as one layer among several. ## Deployment, pricing, and value Initial deployment for a focused use case typically takes four to eight weeks. Timelines depend on the complexity of integrations, the readiness of source content, and the customer's review and approval processes. Larger, multi-department rollouts are usually delivered in phases. Because IdeaBoxAI is a configurable platform rather than custom-built software, most engagements focus on configuration, content preparation, and integration rather than greenfield development. Yes. IdeaBoxAI offers implementation services, and a network of certified partners is available for customers who prefer to work with a regional integrator. Yes. Most successful programs begin with one or two focused Personas in a single department, then expand based on measured results. Routine administration involves updating Personas, refreshing knowledge sources, reviewing usage analytics, and managing access. Most customers operate the platform with a small core team supported by departmental Persona owners. Pricing is structured around platform subscription tiers with capacity for users, Personas, and workload. Costs associated with underlying model usage are tracked transparently and may be billed through IdeaBoxAI or directly from the customer's chosen model provider, depending on the deployment. Detailed pricing is provided during commercial discussions. Cost controls include per-Persona budgets, model routing that directs simpler tasks to lower-cost models, retrieval that limits context size, and caching of frequent queries. Administrators can view consumption by Persona, team, and time period. ROI is typically measured against the specific use case โ€” for example, time saved per ticket, reduction in document review hours, faster proposal turnaround, or fewer escalations. The platform provides usage and outcome metrics to support these measurements. No. Most customers operate IdeaBoxAI with a lean platform team. Departmental users with appropriate training can manage their own Personas under the policies set by the central team. The platform provides analytics on Persona usage, task completion, user feedback, and outcome metrics. Customer success teams help interpret these metrics and identify opportunities for improvement. Personas are designed to be updated as roles evolve. Administrators can revise guidelines, swap or add knowledge sources, and adjust Skills without redeploying the underlying platform. Most customers aim to provide each employee with reliable, role-aware AI assistance that respects company data, follows internal policies, and produces measurable productivity gains over time. ## Persona architecture and customization A Persona is defined by four configurable components: a role definition and system prompt, one or more knowledge sources, a set of Skills (permitted actions), and a layered set of operational rules. Each component has its own management interface. Yes. The platform includes templates for common roles such as customer service, HR, IT support, sales enablement, and research. Templates can be cloned and modified, with shared base policies preserved across derivatives. Yes. Persona configurations are versioned. Administrators can view change history, compare versions, and roll back to a previous configuration if needed. Yes. A Persona can be configured to invoke other Personas for delegated tasks. The orchestrator manages the handoff and enforces each Persona's permissions independently. Skills are typically built on standard API connectors. The platform includes pre-built connectors for common enterprise systems and supports custom Skills built against REST, GraphQL, and other interfaces. Yes. Administrators can configure how much retrieved content and prior conversation is included in each request, which affects both response quality and cost. Yes. Configurations, including prompts, knowledge source bindings, Skill definitions, and policies, can be exported in a structured format for governance review or migration. ## Knowledge, retrieval, and data handling Long-term memory uses retrieval-augmented generation. Source content is ingested, parsed, divided into chunks, embedded into a vector representation, and indexed alongside metadata. At query time, relevant chunks are retrieved and supplied to the language model as grounded context. Both patterns are supported. Knowledge sources can be designated as shared across the organization or scoped to a specific user, team, or Persona. Access controls are applied at retrieval time. Connectors for platforms such as Slack, Microsoft Teams, Confluence, and Jira parse content, preserve relevant structure, and apply metadata such as author and timestamp before indexing. Customers control which sources are included. The platform uses hybrid retrieval that combines semantic (vector) search with lexical (keyword) search, then re-ranks results to maximize relevance. Yes. A strict-grounding mode requires the Persona to answer only when supporting sources are retrieved and to respond with an explicit "insufficient information" message otherwise. This significantly reduces, though does not eliminate, hallucination risk. Documents can carry effective and expiration dates, classification tags, and freshness scores. Retrieval policies can deprioritize or exclude stale content automatically. Yes. Administrators and authorized users can remove specific documents or chunks, and the corresponding vectors are deleted from the index. The platform uses structure-aware chunking that respects headings, paragraphs, tables, and other document elements rather than splitting purely by character count. Multiple chunking strategies are available and can be selected per source. ## Orchestration and workflow execution The orchestration layer coordinates the steps required to fulfill a request: routing the request to the appropriate model, retrieving supporting content, invoking Skills, applying policy checks, and assembling the final response. It also handles retries, timeouts, and fallback logic. Yes. Personas can perform sequences of steps that involve retrieval, reasoning, Skill invocations, and human-in-the-loop checkpoints. Workflow definitions can be deterministic, adaptive, or a combination. Skill selection uses a combination of explicit configuration and model-driven function calling. Each Skill is defined with a schema, and the platform enforces eligibility rules before any Skill is invoked. The orchestrator detects errors and applies configurable retry, backoff, and fallback policies. Persistent failures are reported to the user in a clear way and logged for operations teams. Yes. The platform supports adaptive workflows where the next step is selected based on the current state of the task, in addition to fixed sequences for cases that require deterministic behavior. Inter-Persona communication occurs entirely within the platform's secure environment. Each Persona retrieves only the data it is authorized to access, even when called by another Persona. Yes. Outbound webhooks and integrations with enterprise automation platforms allow workflows to fire events to other systems such as ticketing or alerting tools. Yes. The platform includes a visual builder for designing and reviewing workflows, in addition to a configuration-as-code option for teams that prefer to manage definitions in source control. ## Enterprise security and compliance architecture Every request is authenticated, authorized against Persona and policy bindings, evaluated by content controls, and logged. Only requests that pass these checks reach the model. Outputs are similarly evaluated against output policies before being returned. Yes. The platform can apply detection and masking to prompts, retrieved content, and stored records, using configurable categories and patterns. Detection accuracy depends on the data; customers should validate against representative samples. Identity attributes from your identity provider are mapped to Persona access policies. Users see only the Personas they are entitled to and, within each Persona, only the knowledge and Skills permitted by their role. Yes. Retention windows are configurable per data category โ€” for example, conversation history, audit logs, retrieved content, and Persona memory โ€” to align with the customer's policy and regulatory obligations. Detailed audit records cover authentication, authorization decisions, configuration changes, Persona interactions, Skill invocations, and policy enforcement events. Audit data can be exported to customer SIEM systems. Yes. On-premises deployments support fully disconnected operation when paired with a self-hosted model, suitable for sensitive environments with strict data egress restrictions. Customers can select regions for storage and processing in supported deployment models. For on-premises and customer-managed cloud deployments, data residency is determined by the customer's infrastructure. ## Models, FinOps, and performance Model capabilities, pricing, and availability change frequently. A model-agnostic platform lets customers select the most appropriate model for each task and adjust over time without rewriting Personas or workflows. Supported providers typically include OpenAI, Anthropic, Google, Microsoft Azure OpenAI, AWS Bedrock, and self-hosted open-source models such as Llama and Mistral variants. The current list is maintained in the product documentation. Routing is rule-based and configurable. Administrators can set defaults per Persona, route specific task types to specific models, and define fallback chains for availability and cost optimization. If a fallback model is configured for a Persona, the orchestrator routes traffic automatically. Customers are notified of degraded modes when the fallback differs materially from the primary model. Yes. Customers can connect a self-hosted model endpoint, including open-source models running in their own environment, to keep model inference fully within their infrastructure. Standard metrics include time to first token, end-to-end response latency, retrieval relevance, task success rate, and user feedback scores. Dashboards present these metrics by Persona, team, and time period. By supplying only the most relevant content to the model rather than entire documents, retrieval reduces input token usage and often improves response quality at the same time. Yes. Response caching is supported with configurable scope and time-to-live. Cached responses respect access controls so that users see only content they are entitled to. Adding retrieval, policy evaluation, and logging introduces a modest amount of overhead. In most enterprise workloads, this overhead is small compared with model inference time and is offset by improved accuracy and governance. ## Advanced topics and differentiation Customers can certainly build elements of an AI platform in-house. IdeaBoxAI's value is in providing the integrated combination of orchestration, retrieval, governance, integrations, and lifecycle tooling โ€” together with ongoing maintenance โ€” without each customer having to build and maintain it independently. The platform wraps language models with a deterministic policy and orchestration layer. Where business logic must be reliable โ€” for example, applying eligibility rules or routing approvals โ€” that logic is enforced outside the model, with the model used for the language tasks it does best. Yes. Integrations with systems including Salesforce, Microsoft Dynamics, SAP, Workday, ServiceNow, Zendesk, and others are available through Skill connectors. Custom integrations can be built using standard API patterns. Yes. Tenant isolation features allow service providers to maintain separate Personas, knowledge, and configurations per client within a single deployment. Yes, for customers who embed IdeaBoxAI into their own products. White-labeling options are available subject to commercial terms. Implementation playbooks recommend starting with high-quality seed content such as existing standard operating procedures and curated examples. Iterative tuning and feedback collection during a pilot phase further improve performance before broad rollout. Image inputs are supported with multimodal models. Voice and other modalities are supported on an evolving basis as the underlying model capabilities mature. Current support is documented in the product release notes. Customers typically present ROI in terms of time saved on defined tasks, reduction in error or rework, faster cycle times, and avoided spend on point solutions. The platform provides metrics that map naturally to these financial measures. ## Architecture and infrastructure IdeaBoxAI uses a service-oriented architecture composed of independently scalable components for ingestion, retrieval, orchestration, model routing, governance, and API serving. Services are typically containerized and orchestrated with Kubernetes in production deployments. Three primary models: multi-tenant SaaS, single-tenant private cloud in the customer's AWS, Azure, or Google Cloud account, and on-premises deployment for environments with strict residency or air-gap requirements. Yes. Reference Terraform and Helm configurations are provided for customer-managed deployments to enable repeatable, version-controlled provisioning. Stateless services are scaled horizontally. Persistent state โ€” including Persona configuration, audit logs, and vector indexes โ€” is stored in managed databases and object storage suited to each data type. Yes. Production deployments are designed for high availability across multiple availability zones with automated failover for critical components. Long-running and asynchronous workflows are managed by a task queue and worker pool, with status tracking exposed through the API and user interface. Event-driven integrations are supported through standard message brokers and managed cloud queue services. The platform can both produce and consume events. Yes. Common hybrid patterns include keeping the control plane in the cloud while indexing and retrieving from data sources kept on the customer's network, or vice versa. Service-to-service communication uses standard protocols including HTTPS and gRPC, with mutual TLS available for sensitive paths. Logs are structured (typically JSON) and compatible with OpenTelemetry. They can be exported to common observability platforms such as Splunk, Datadog, Elastic, and cloud-native logging services. Schema migrations are managed through automated pipelines using standard tooling, with backward-compatible deployments to support rolling upgrades. In multi-tenant deployments, isolation is enforced at the application, data, and network layers. Tenant identifiers are validated on every request, and data stores are partitioned or scoped per tenant. Container images are built for multiple architectures where supported, including ARM-based options such as AWS Graviton, to give customers flexibility in cost and performance. Service discovery uses the orchestration platform's native mechanisms (for example, Kubernetes services). Customers can integrate with service mesh tooling where their environment standardizes on one. Resource requirements vary by deployment size and chosen options. A baseline reference is provided in the technical documentation, and the IdeaBoxAI solutions team helps size deployments based on expected workload. ## Data architecture and RAG internals Supported options typically include managed services such as Pinecone, as well as self-hosted engines such as Milvus, Weaviate, Qdrant, and pgvector. The current list is maintained in the technical documentation. Yes, where the schema and embedding model can be aligned with IdeaBoxAI's ingestion framework. The solutions team can advise on compatibility for specific cases. Embeddings are produced by a configurable embedding model at ingestion time and updated as source content changes. Multiple embedding model providers are supported. Strategies include fixed-size chunking, structure-aware chunking that follows document headings and tables, and semantic chunking that groups related sentences. The strategy can be selected per source. The ingestion pipeline uses document understanding tools to extract tables in a structured form, preserving relationships between rows and columns. Figures and images can be processed by multimodal models where applicable. Yes. Hybrid retrieval combines dense vector search with lexical search and applies a re-ranking step to improve final relevance. Each response can include citations to source documents, and an internal trace records the retrieval, model, and policy decisions that produced the answer. Yes. Metadata tags such as department, classification, date, and source can be used to filter retrieval before vector similarity is applied. Connectors poll or subscribe to source systems on a schedule appropriate to each source. Near-real-time updates are supported for systems that provide change notifications. The ingestion pipeline removes the corresponding chunks from the index in line with the source system's deletion event, subject to retention policy for audit purposes. Re-ranking and ordering strategies place the most relevant content where the model is most likely to use it. Customers can tune these strategies per Persona. Yes. Multi-stage retrieval โ€” for example, identifying the relevant document first, then drilling into specific sections โ€” is supported for very large corpora. Connectors strip extraneous formatting, preserve thread and authorship metadata, and apply consistent structure so retrieval behaves predictably across different source types. Knowledge graph integration is supported for use cases that benefit from explicit entity and relationship modeling. Graph results can be combined with vector retrieval in hybrid patterns. Infrequently used content can be tiered to lower-cost storage with on-demand rehydration into the active index when accessed. ## Integration and API Through a documented REST API and language-specific SDKs. Embeddable UI components are also provided for common scenarios such as in-app chat experiences. REST is the primary API style. GraphQL endpoints are available for specific use cases; the current scope is maintained in the developer documentation. Skills are exposed through a unified schema. The platform translates between this schema and the function-calling conventions of each supported model provider so that customers do not have to maintain provider-specific code. Yes. The orchestrator can execute independent Skills in parallel when the workflow permits and assemble the results before producing a response. Public endpoints support OAuth 2.0 client credentials flow, OpenID Connect, and API keys, with optional IP allow-listing and mutual TLS. Internal service-to-service communication uses mutual TLS where appropriate. Yes. Skill connectors are available for major enterprise systems, and custom connectors can be developed against published APIs. The platform applies request shaping, retries with exponential backoff, and circuit breakers to protect downstream systems and provide graceful degradation. Yes. Customers can integrate IdeaBoxAI with platforms such as MuleSoft, Workato, and Boomi to centralize integration logic where appropriate. Webhooks fire on configurable events. Payloads are signed using HMAC so that receiving systems can verify authenticity. Yes. Skills can wrap SOAP services, with the platform translating between modern interfaces and the underlying SOAP contract. Large files are uploaded through pre-signed storage URLs and processed asynchronously by background workers. Status is exposed to the user during processing. Yes. Streaming responses are supported via standard mechanisms such as Server-Sent Events, suitable for interactive chat experiences. Yes. Customers can add custom logic at defined extension points where appropriate. The platform documents supported extension patterns and their lifecycle implications. The API uses versioned endpoints with a published deprecation policy. Significant changes are communicated in advance, and parallel versions are supported during transition periods. Yes. Customers can use a sandbox environment for development and testing without affecting production data or workloads. ## Identity, zero trust, and threat defense The platform assumes no implicit trust between components. Each request is authenticated, authorized, encrypted in transit, and logged. Network segmentation, least-privilege roles, and continuous verification are core design principles. AES-256 for data at rest, TLS 1.2 or higher for data in transit, and support for customer-managed keys via AWS KMS, Azure Key Vault, or Google Cloud KMS on applicable deployment options. The platform acts as a service provider with SAML 2.0 and OpenID Connect, integrating with identity providers such as Microsoft Entra ID, Okta, Ping, and Google Workspace. Passwords are not stored locally for SSO users. Yes. Access policies can use attributes from the identity provider โ€” for example, department, role, or region โ€” to control which Personas, knowledge sources, and Skills a user can use. Defenses include input validation, isolation of system instructions, output policy enforcement, and tool-use authorization checks. The platform follows current industry guidance, which continues to evolve. Yes. Customers can integrate with their DLP tooling to scan prompts and outputs against organizational policy. Configurable detectors identify common PII categories and apply masking, redaction, or blocking according to policy. Customers can extend detection with custom patterns. Administrative access integrates with privileged access management tooling, and just-in-time elevation patterns are supported on applicable deployments. Credentials are stored in encrypted form using the platform's secrets manager and can be integrated with external vaults such as HashiCorp Vault. Plain-text storage of credentials is not used. Network-level controls such as IP allow-listing and regional restrictions can be configured for the control plane and APIs. Container images are scanned for known vulnerabilities, signed, and built to run with least privilege. Hardening guidance is provided for customer-managed deployments. Customers can run their preferred runtime security agents alongside the platform in customer-managed deployments and forward telemetry to their security operations tooling. Yes. Role separation between infrastructure administrators, security administrators, and Persona administrators is supported through configurable roles. Customer-managed encryption keys, opted-out training settings, scoped credentials, and contractual commitments from model providers limit exposure. On-premises and self-hosted model deployments remove the upstream provider entirely from the data path. All configuration changes and privileged actions are recorded in an immutable audit log with user, time, and change details, and can be exported to a customer SIEM. ## Model management and MLOps Routing is configurable. Defaults are set per Persona, and rules can route specific task types or user segments to specific models based on capability, cost, and latency. Yes. Self-hosted endpoints โ€” including open-source models โ€” can be registered with the model gateway and used by Personas in place of, or alongside, hosted providers. Primary and secondary models can be configured per Persona. On failure, timeout, or capacity issues, the orchestrator routes to the configured fallback with logging. Yes. Traffic can be split between models for a Persona, with metrics tracked separately for each variant to support data-driven decisions. The platform tracks token usage against each model's context window and applies retrieval, summarization, and truncation strategies to stay within limits. Metrics include time to first token, tokens per second, total latency, retrieval relevance, task success rate, and user feedback. Metrics are available per Persona, model, and time period. User feedback and outcome metrics are tracked over time so that significant changes in quality can be detected and investigated. Customers receive guidance when underlying models are updated by providers. Yes. Customers can register fine-tuned model endpoints from their providers and direct selected Personas or tasks to them. Prompts and configurations are versioned within the platform and can additionally be exported to source control for organizations that manage them alongside other code artifacts. Yes. Multimodal inputs such as images are supported with capable models, and Persona memory can reference non-text content where appropriate. Quality is supported by retrieval grounding, explicit operational rules, output policies, and user feedback loops that surface low-quality responses for review. Yes. Sampling parameters can be configured per Persona and per task type, balancing creativity and determinism according to the use case. The model gateway queues and shapes requests, applies priorities where configured, and scales worker capacity to absorb demand within provider rate limits. When an embedding model is changed, the platform supports background re-embedding of content with progress tracking to maintain retrieval consistency. Provider credentials are managed centrally in the platform's secrets store and never exposed to end users. The gateway injects them into provider calls at runtime. ## Responsible AI, ethics, and compliance practice The platform provides features that support common obligations for higher-risk AI systems, including documentation of intended use, logging, human oversight options, and configurable controls. Customers remain responsible for their own classification and compliance posture; IdeaBoxAI provides supporting materials and guidance. Yes. Workflows can require human approval before sensitive Skills are executed โ€” for example, before sending a customer-facing message or modifying a record in a system of record. Retrieval-grounded responses tied to reviewed enterprise content reduce reliance on broad model knowledge. Customers are encouraged to test Personas with representative scenarios as part of acceptance and ongoing operation. Yes. Output content classifiers can be applied to flag or block responses that violate organizational policy. Strict-grounding modes limit responses to retrieved organizational content. Customers should still apply their own legal review for content reuse where appropriate. Operational rules can require domain-specific Personas to include relevant disclaimers and to escalate or decline questions outside their scope. Customers configure these behaviors to match their internal standards. Yes. Customers can define rules for retaining or purging PII separately from other categories of stored data. Yes. Immutable audit logs, versioned configurations, and selective retention rules support legal holds and discovery requests. Visible attribution and metadata tagging of AI-generated content are supported. The state of provenance and watermarking standards is evolving, and the platform tracks emerging standards. Responses can include citations to source documents, and a trace view shows which sources, models, and policies contributed to a response. This supports both end-user trust and administrative review. ## Scalability, reliability, and operations End-to-end latency depends on the selected model, retrieval depth, and workflow complexity. Typical interactive queries complete in a few seconds; specific service levels are documented in the customer's agreement. Ingestion workers and retrieval services scale horizontally based on workload. Capacity planning guidance is provided based on corpus size and query rate. Scaling follows the chosen vector database's mechanisms โ€” managed scaling on hosted services, or sharding and replication on self-hosted engines. The platform abstracts these differences from Persona configuration. Repeated failures from a downstream Skill or model open a circuit and route the workflow to a configured fallback path, with automatic recovery once health is restored. Standard pooling mechanisms protect databases from connection exhaustion under load. Defaults are tuned for typical enterprise workloads. Yes. Analytics and reporting queries can run against read replicas to isolate them from transactional workloads. Production deployments include backups, multi-zone redundancy, and documented recovery procedures. Specific RPO and RTO targets are defined in the customer's service agreement. Long-running tasks are handled asynchronously and tracked with status APIs. Clients can subscribe to completion events rather than holding open connections. Background maintenance routines remove orphaned vector entries, expired cached responses, and other transient data on a scheduled basis. Yes. Per-team and per-Persona quotas can be configured to control consumption of model and infrastructure resources. ## Procurement, FinOps, and roadmap Customers can switch model providers, swap vector databases, and export their configurations and data. The platform's value lies in orchestration and governance rather than in a proprietary model or proprietary data format. Customer data โ€” including Persona configurations, knowledge sources, conversation history (subject to retention policy), and audit logs โ€” can be exported in standard formats on request. Exit procedures are described in the customer agreement. Consumption dashboards translate AI usage into financial measures and break costs down by Persona, team, and time period. Budgets, quotas, and alerts help organizations manage spend proactively. Yes. Enterprise customers receive SLAs covering availability of the control plane and API endpoints. Specific service-level commitments and remedies are documented in the customer agreement. Enterprise customers can participate in advisory programs, review forward-looking roadmap information under non-disclosure, and submit prioritized feature requests through a managed intake process. ## Detailed competitive differentiation ### Custom GPTs and consumer assistant patterns Custom GPTs let users package instructions, files, and limited actions into a tailored chat experience within ChatGPT. Personas provide similar packaging and add features common in enterprise environments: granular role-based access control, layered policies, versioning, audit logging, multi-model routing, and integrations governed by IT. Custom GPTs rely primarily on uploaded files and conversation context within ChatGPT. Personas use a retrieval architecture that can scale to much larger corpora and apply per-source access controls. No. Personas are scoped to the customer's organization. Some platform components and templates are shared, but customer-configured Personas remain inside customer tenants. Both products use user prompts as the primary input. Personas additionally support structured triggers, scheduled execution, and event-driven workflows, which means a Persona can act without a user typing a prompt. Custom GPTs typically require re-uploading or refreshing files. Personas can be wired to live knowledge sources through connectors, so updates propagate without manual reuploads. Custom GPTs use a single instructions field. Personas separate behavioral guidance, operational policies, and Skill rules so that conflicts are minimized and policies are easier to maintain. Yes. Because Personas are model-agnostic, switching the underlying model is a configuration change rather than a rebuild. Custom GPTs can call Actions and tools to perform multi-step work; Personas add orchestration features such as conditional branching, parallel Skill execution, and human-in-the-loop checkpoints. Persona creation is governed by configurable workflows. Organizations can require IT or platform team approval before a new Persona is published to end users, reducing unmanaged AI usage. Yes. Personas can be assigned service identities for system-to-system interactions, with all activity logged against that identity for traceability. Persona workflows include explicit error handling, retry policies, and fallback paths so that failures during multi-step tasks are managed rather than surfaced as raw errors. Consumer assistant products are primarily chat-centric. IdeaBoxAI is task-centric: chat is one interface among several, and many Personas operate entirely through APIs and scheduled triggers. System prompts are kept out of user-visible contexts where possible, and output policies block requests to disclose them. As with any AI system, defense in depth is needed; sensitive policy should not rely solely on prompt secrecy. Yes. The orchestration layer supports controlled delegation between Personas, with each Persona's permissions enforced independently. Because it separates the design and approval of an AI assistant from its day-to-day use, which is what most large organizations require for any production system. ### Project-style knowledge spaces Project-style workspaces typically group documents and instructions for a particular topic. Personas use a broader retrieval architecture with role-based access control on each source, metadata-driven filtering, and re-ranking, designed to scale to organization-wide knowledge. Persona retrieval is semantic and hybrid by default and applies metadata filters and access controls before similarity search, which reduces the chance of retrieving content the user is not entitled to see. Through metadata-driven scoping, hybrid retrieval, and re-ranking. Personas can also be limited to specific knowledge bundles per task to avoid cross-contamination of context. No. Retrieval supplies only the most relevant content to the model, which is generally more reliable and cost-effective than relying on extremely long contexts. Yes. Document versions, effective dates, and update history can be tracked and used to prioritize current content during retrieval. Connectors normalize formatting and structure during ingestion, so retrieval behaves consistently across diverse sources. Yes. Through Skills, a Persona can query live systems in real time for data that should not be cached or pre-indexed. Persona use spans informational tasks such as summarization and Q\&A as well as transactional tasks that update enterprise systems through Skills. Permissions are enforced at retrieval time using the user's identity and attributes, so two users asking the same question may see different supporting content based on their entitlements. Personas can be configured to produce structured outputs such as JSON conforming to a defined schema, in addition to natural-language responses. Tiered storage, scheduled re-indexing, and selective archiving keep retrieval responsive as content grows. Connectors update indexes on a configurable cadence, so Personas reflect current content without manual intervention. Sources can be tagged with reliability and category metadata, and Personas can be configured to prefer authoritative sources for factual answers. Multimodal inputs are supported with capable models, and content metadata can be used to apply policy to non-text material. Yes. Adds, removes, and modifications to knowledge sources are tracked, supporting governance review and rollback. ### Productivity suite copilots Suite-bound copilots primarily access data within their parent suite. IdeaBoxAI is application-agnostic and integrates with a broad set of enterprise systems beyond any single productivity suite. IdeaBoxAI supports embedding into existing tools โ€” including Microsoft Teams, Slack, custom intranets, CRM interfaces, and mobile apps โ€” as well as a standalone web interface. Suite-bound copilots focus on creating documents within their suite. Personas can produce documents as part of broader workflows that include data retrieval, system updates, and approvals. Tools are integrated through the Skills framework with explicit schemas, permissions, and auditing, which is generally more flexible than plug-in patterns tied to a single product. Personas apply policy and access controls per source, separating informal communications from formal documents and other content categories. Personas can model deep, domain-specific roles with their own policies, knowledge, and Skills, beyond what general-purpose copilots typically allow. Customers can run IdeaBoxAI on the cloud of their choice or on-premises, which is helpful for organizations with diverse infrastructure footprints. Persona memory is persistent and grounded in your governed knowledge sources rather than reliant on session-only context. The orchestration layer plans, executes, and verifies multi-step work, including conditional branches and approvals where required. Yes. Policies apply uniformly to users regardless of seniority, so Personas can decline actions that violate compliance or operational rules. Connectors are first-class. Adding a new data source is a configuration task rather than a custom development project. Responses can be tied to retrieved sources, and strict-grounding modes are available where it is important to avoid responses unsupported by retrieved content. Self-hosted, air-gapped deployments are supported for use cases that require fully local operation. ### Low-code assistant builders Low-code canvases typically use deterministic flows. IdeaBoxAI supports adaptive workflows that select the next step based on context, while also supporting deterministic flows where they are appropriate. Through native Skills with explicit schemas and authentication, rather than wrapper integrations that delegate to a separate automation product. The orchestrator can interpret intent, ask clarifying questions, or escalate to a human, rather than failing when the input does not match a predefined path. Most updates are configuration changes โ€” to policies, knowledge, or Skills โ€” rather than redesigns of a visual canvas, which can shorten iteration cycles. Both. Personas can serve as conversational assistants and as autonomous task executors triggered by events or schedules. A trace view exposes the inputs, retrievals, model selections, and Skill calls that produced each response, which helps with troubleshooting and audit. Yes. The same Persona can use different models for different task types, managed centrally by the model gateway. The platform retries, queues, and falls back according to policy so that downstream rate limits do not cause user-visible failures whenever avoidable. Through configuration updates informed by feedback and analytics. Foundational model training is the responsibility of model providers. Personas with capabilities, rather than topics with rigid branches, which scales better for complex enterprise scenarios. Centralized policies apply to all Personas by default, so individual designers do not need to recreate compliance rules in each workflow. Defined extension points support custom code in supported languages, and APIs and SDKs let developers integrate IdeaBoxAI deeply with their own systems. Persona configurations explicitly define how context is preserved, scoped, or reset between tasks. Many Personas are deployed in days to a few weeks. Larger, multi-system workflows typically take longer based on integration scope and review processes. Yes. Personas can run on schedules or in response to events and deliver output through email, messaging, or system integrations rather than chat. ### Administration and governance comparisons Analytics emphasize business outcomes โ€” tasks completed, time-to-resolution, deflection rates, satisfaction โ€” alongside technical metrics like token usage and latency. Retention rules can be set per data category, with scheduled enforcement and immutable audit trails of changes. Controls extend down to individual Personas, knowledge sources, and Skills, with attribute-based policies applied at runtime. Through immutable, exportable audit logs that record relevant authentication, authorization, configuration, and usage events. Identity attributes from your identity provider map to Persona permissions, so changes in your IdP propagate to IdeaBoxAI. Network controls can restrict access to the control plane and API endpoints to specified IP ranges or via secure tunnels. Yes. Per-Persona and per-team budgets are configurable, with alerts and optional automatic throttling when limits are approached. Through input policy enforcement, output filtering, retrieval grounding, and ongoing monitoring. As with any AI system, no defense is perfect. Customers can pin Personas to specific model versions where supported, test changes in staging, and roll out updates in a controlled manner. Yes, where Personas are surfaced in customer-owned interfaces. The level of branding depends on whether the Persona runs in the IdeaBoxAI UI or is embedded in a customer application. Bulk policy updates, exports, and migrations are supported through APIs and management tools to make large fleets of Personas operable by small teams. Through contractual commitments with model providers, configuration of training opt-outs, and the option to use self-hosted models that remove the question entirely from the data path. Logging policies can mask or hash sensitive fields, so administrators can troubleshoot without exposure to underlying personal data. Yes. Central administrators can set platform-wide policies while delegating specific build and operation rights to business units. Identity providerโ€“driven deprovisioning revokes access, and configurable rules can purge or anonymize user-specific data on offboarding. ### Responsible use and output control Through configurable operational rules. Refusals are based on documented organizational policy, with an option to redirect users to alternative resources. Personas can be configured for specific response styles and structured outputs, including strict JSON schemas where downstream systems require them. It responds explicitly that it does not have enough information, optionally citing what it did find, and can route the question to a human or another Persona if configured to do so. Accurate, grounded, policy-aligned responses that are useful in a business context. Personas are designed to defer rather than guess when supporting evidence is missing. Yes. Lexicon and style guidelines can be applied, and output policies can enforce required disclosures or block disallowed terminology. Through a documented priority order: organizational policy first, then Persona configuration, then user instructions. The resolution is visible in audit traces. Yes. Specific Skills or stages can require human approval before execution, with the request and decision recorded in the audit trail. Yes. Through Skills, Personas can update systems, send communications, and trigger workflows in addition to responding to questions. Personas can be configured with locale-specific guidelines, terminology, and disclosures. Language support depends on the underlying model and on retrieval coverage in each language. Through domain-specific policies and feedback-driven tuning, so business content is not unnecessarily blocked while genuinely problematic content is still managed. ### Outputs and execution Outputs can be displayed in chat, exported to documents, written into target systems through Skills, or pushed to repositories โ€” depending on the workflow. Yes. Skills allow generated content to be used as input to updates in CRM, ERP, ITSM, and other systems, subject to approvals and policy. Yes, through standard mechanisms appropriate to the output type, including links, exports, and integrations with collaboration tools. Outputs that matter for audit or compliance can be stored with version metadata, with history accessible through the platform's interfaces. Yes. Workflow events can trigger notifications in messaging tools or operational systems. ### Model strategy in practice Model availability, pricing, and capabilities change. A platform that supports multiple providers lets customers adapt without rewriting their AI workflows. Yes. Routing rules can direct different task types within a Persona to different models โ€” for example, a fast model for short clarifications and a more capable model for complex analysis. Tasks can be routed to the most cost-effective model that meets the latency and quality requirements, often producing meaningful savings versus using a single premium model for everything. Yes. Traffic-splitting and outcome tracking let customers make data-driven choices about which model best fits a given task. Configured fallbacks reroute traffic automatically, with logging and customer-visible status. Token counts are recorded per request, aggregated by Persona, team, and model, and exposed through dashboards and exports. Yes. The platform can apply model-specific prompt adjustments to maintain consistent behavior across providers. IdeaBoxAI is intentionally model-neutral. It is designed to let customers benefit from advances across the foundation model market while keeping their orchestration, governance, and integrations stable. Yes. The supported model catalog is maintained in the technical documentation and updated as new models are validated. # Frequently asked questions Source: https://docs.ideaboxai.com/support/faq Learn the answers to the most common questions about IdeaBoxAI accounts, features, and troubleshooting. Find answers to the most common questions about getting started with IdeaBoxAI, managing your account, and using the platform features. ## Account and access ### How do I create an IdeaBoxAI account? Go to [app.ideaboxai.com/auth/login](https://app.ideaboxai.com/auth/login) and click **Sign up**. You can register with an email and password or use Google or Microsoft SSO. See the [quickstart guide](/quickstart) for a full walkthrough. ### What SSO providers does IdeaBoxAI support? IdeaBoxAI supports Google and Microsoft SSO. Click the relevant icon on the login screen to authenticate with your existing account. ### How do I reset my password? On the [login page](https://app.ideaboxai.com/auth/login), click **Forgot Password?** and enter your email address. You will receive a reset link in your inbox. ### Can I invite team members to my workspace? Yes. Workspace admins can add team members and assign them to personas from the Admin Console. Each user logs in via SSO and sees their role-specific Persona AI immediately. ## Persona AI and personas ### What is a persona? A persona is the AI's understanding of your role, your goals, and the data sources you care about. When you log in, the Persona AI tailors its responses to your specific function. Learn more in the [Persona AI introduction](/copilot/introduction). ### Can I create a custom persona? Yes. In addition to the built-in personas, your workspace admin can create custom personas tailored to any role. Each custom persona has its own skills, knowledge bases, and starter prompts. Contact your admin to request a new persona for your team. ### What data sources does the Persona AI connect to? The Persona AI connects to Salesforce, your data warehouse, Google Drive, Confluence, and other sources configured by your admin. All responses are grounded in your actual business data, not generic web results. ## Agentic BI and dashboards Beta ### Do I need SQL knowledge to build dashboards? No. The dashboard builder lets you ask questions in plain English. IdeaBoxAI generates SQL-backed visualizations from your natural language queries. ### What types of visualizations are supported? IdeaBoxAI supports auto-generated charts, KPI cards, and interactive visualizations with drill-down capabilities. See the [Agentic BI introduction](/agentic-bi/introduction) for details. ## Troubleshooting ### I am not receiving the verification email Check your spam or junk folder. If the email does not arrive within a few minutes, then try registering again or contact [support@ideaboxai.com](mailto:support@ideaboxai.com). ### The Persona AI is not returning data from my CRM Verify that your data source is connected in the Connections settings. If the connection shows as active but data is missing, then contact your workspace admin to check permissions. ### How do I contact support? Email [support@ideaboxai.com](mailto:support@ideaboxai.com) for any issues not covered in this FAQ. # Release notes Source: https://docs.ideaboxai.com/support/release-notes Stay up to date with the latest IdeaBoxAI platform updates, new features, and improvements. Check this page for a running history of what's new on the IdeaBoxAI platform, from major feature launches to smaller improvements. This release adds sharing and version history across the Persona AI platform, a new Drive to hold every file your personas generate, and workspace branding controls for admins. ### Sharing and collaboration **Share a conversation.** Generate a link to any Persona AI chat session from the Share button at the top of the chat window. The recipient can view the full conversation and continue chatting from exactly where it left off. Shared links are time-limited and the exact expiry is shown to you at the moment you share. Share conversation dialog showing a generated link with its expiry time and a Copy link button. **Share a persona.** Control exactly who can use a persona you've built, with three visibility levels: * **Private** (default): only you can access the persona. * **Organization**: anyone in your workspace can access and use it. * **Public**: anyone with the link can access it. Share persona dialog with the Private access level selected and a disabled Copy link field. **Share a skill.** Generate a share link for any skill directly from the Skills Studio. When a teammate opens the link, they choose which project and persona to add the skill to, making it easy to hand a skill to another team without rebuilding it. Skills Studio card for a skill with the Copy share link action highlighted. ### Versioning **Persona version history.** Every edit to a persona now creates a new version automatically. Open the version history panel to see each past version, review what changed, and promote any earlier version back to active, without losing your current configuration. Persona version history panel showing the current active version and its description. **Skill version history.** Skills follow the same versioning model as personas. The version history panel lists every past version along with its change description, so you can track how a skill has evolved and roll back if needed. Skill version history panel showing the current active version and its description. ### Drive A new **Drive** section in the sidebar acts as a central store for every file your Persona AI generates. Each file links back to the session that created it, so you can always trace a file to its original conversation. * **File management**: download, share via a copy link, or delete any file straight from Drive. * **Edit in Persona AI**: reopen any Drive file and continue editing it by launching it back into a Persona AI session. * **Session navigation**: jump from any file card back to the conversation that generated it. * **Grid and table views**: browse your files as a visual grid or a sortable table, with keyword search and file-type filters. * **File versioning**: when a file is regenerated or updated, Drive automatically keeps the new version accessible from the file's detail page. The Drive page, showing where every file your Persona AI generates will appear. ### Dashboard * **Move artifacts to your Dashboard**: any HTML artifact your Persona AI generates (a chart, a report, an analysis page) can be moved straight to your Dashboard from the artifact preview, giving you one dedicated place to collect your key outputs. * **Pin your favorites**: pin the dashboards you use most often to the top of the list for faster access. The Dashboards page where HTML files moved from Drive can be pinned to the top of the list. ### Admin Console: branding Workspace admins can now customize how the workspace and Persona AI look, from a new Branding section in the Admin Console: * **Identity**: set the workspace name, upload a logo (PNG or JPEG, up to 2MB), and configure a favicon for browser tabs. * **Colors**: customize primary and accent color palettes using a color picker or hex input. * **Typography**: choose the workspace font from the available options. * **Live preview**: every branding change updates instantly in a live preview panel before you save. * **Reset to default**: restore the default IdeaBoxAI theme at any time. Admin Console Branding page showing workspace name, logo upload, colour palettes, typography, and a live preview panel. This release introduces IdeaBoxAI Persona AI, a persona-driven AI assistant that adapts to your role, along with a redesigned navigation, an integration framework connecting 1,000+ tools, and a guided self-service onboarding flow. ### IdeaBoxAI Persona AI **Persona AI core** * **Landing page and chat interface**: a dedicated landing page shows your recent session history with one click to start a new chat. Getting-started prompts are personalized based on your own prior interactions. * **Persona system**: every user is assigned a role-specific persona that shapes Persona AI's behavior, skills, and available workflows. You can switch personas mid-session. * **Skills and knowledge**: equip your persona with skills and connect knowledge bases and integrations, so Persona AI can take actions, retrieve context, and generate outputs tailored to your role and your data. * **Memory**: Persona AI remembers context across sessions for each persona, so it becomes more useful the more you use it. * **Chat history management**: every conversation is stored and searchable, with options to pin, rename, and delete sessions. Each session links back to the individual messages within it. * **Model switching**: switch between supported AI models directly from the chat interface. The active model is always shown in the chat input bar. **Platform skills** * **Web search and browse**: Persona AI can search the web and read live web pages to bring in current information and external context in real time. * **Document and artifact generation**: using built-in sandboxed execution, Persona AI can produce and download richly formatted documents from the chat: HTML reports, PowerPoint presentations, Word documents, and PDFs, without leaving the conversation. * **AI image generation**: generate visuals directly from the chat interface using the dedicated image-generation model option. * **File and image attachments**: attach files (like PDFs and documents) and images (PNG, JPG, and more) to a conversation to give Persona AI extra context. * **Skill creation via chat**: describe a new skill directly in the chat, and Persona AI creates it and assigns it to your active persona automatically, with no manual configuration needed. **Admin Console and user settings** * **Admin Console**: a full management surface for admins, covering user management, org-level connection approvals, observability, role-based access control, and an audit log with CSV export. * **User settings**: self-service profile management, including your personal integrations and usage activity. ### Navigation and UI revamp The sidebar has been redesigned around three clear groups: **Menu** (core product areas), **Settings** (configuration and profile), and **Support** (help and documentation), with content organized by project so everything tied to a workspace is easier to find. * Personas and Skills have moved out of the Admin Console and into the main Menu, so any user can create and manage them without needing admin access. * Connections has moved from a nested tab under Settings to a first-class item in Settings, making it faster to reach. ### Integration connectors A new integration framework lets Persona AI read from and write to the external business tools you already use, with authentication handled transparently. * **Connection management**: browse available connectors, connect, view connection status, and disconnect, all from a dedicated Connections page. Connections are managed per user. * **1,000+ connectors**: access over 1,000 tools and services out of the box, spanning CRMs, productivity suites, communication tools, data platforms, and more. * **Dynamic connector registry**: new connectors can be added over time without requiring you to do anything on your end. ### Freemium onboarding A guided, self-service setup flow takes new users from sign-up to a fully configured, personalized workspace with no admin required. * **Sign-up and onboarding wizard**: a multi-step wizard tracks your progress and resumes from the last step if you drop off and come back later. * **About you**: tell us your company name, website, functional area, and a short business description. Everything auto-saves as you go, and an "Improve with AI" option can rewrite your description for you (with the option to accept or revert). * **Persona selection**: see an AI-generated summary of your company alongside a ranked set of suggested personas matched to your context. Pick one, or create your own custom persona. * **Data setup**: connect a live data source, or start with AI-generated sample data seeded from your company context, while your workspace is prepared. * **Workspace launch**: watch your workspace get built in real time, then land in it fully configured, complete with an interactive in-app tutorial. * **Business DNA**: after launch, refine your business context at any time and regenerate persona suggestions to match. * **Credits and wallet**: freemium users receive 10,000 credits per week, with in-app alerts as you approach your limit. ### Help Center * **In-app Help Center**: get guidance on using the platform without leaving it. Answers are backed by search over IdeaBoxAI's documentation, with citations back to the source articles. * **Feedback and bug reporting**: submit feedback or report a bug directly from the Help Center. ### Usage and observability * **Your usage**: see your own credit consumption and model-level usage breakdown, with 7-, 30-, and 90-day views. * **Admin usage view**: admins can see the same data aggregated across the whole workspace. ### Knowledge Base The Knowledge Base creation flow has been redesigned for a smoother, more guided experience when adding, indexing, and managing your knowledge sources. ### Dashboard experience (beta) A full refresh of the dashboard list and detail pages for a faster, more interactive experience. All of your existing dashboard workflows (viewing, editing, renaming, deleting, and theming) continue to work exactly as before. * **Refreshed list page**: a sticky search toolbar, sortable columns, and a List/Grid view toggle, with an inline actions menu on every card for Open, Rename, Duplicate, and Delete. * **Dashboard metadata at a glance**: see the dashboard name, creation date, and creator directly in the list view. * **Grid view with thumbnails**: switch to Grid view to see a visual preview of each dashboard's chart layout. * **Inline duplicate, rename, and delete**: manage dashboards directly from the list, with a confirmation step before anything is deleted. * **Last refreshed timestamp**: always see when a dashboard's data last updated. * **Refresh scheduling**: configure an automatic refresh schedule, or trigger a manual refresh on demand. * **Chart hover actions**: hover over any chart for quick access to copy, edit, delete, and chart-type options. * **AI-assisted chart editing**: open the Dashboard Agent from any chart to describe a change in natural language: a new filter, a different chart type, a rename, and have it applied without leaving the page. This release makes it easier to explore data conversationally and to understand what your charts are actually telling you. **Navigate data hierarchies with one click.** Roll up and down through levels like year, quarter, and month. The whole chart transitions between levels at once, without needing to select individual data points. A level indicator always shows you where you are (for example, "Quarter ยท Level 2 of 3"), and your active global filters stay applied at every level. **Build and edit dashboards by chatting.** Describe the dashboard you want in plain language, and the Dashboard Bot handles the rest: discovering the relevant data, generating titles, layouts, KPI cards, and charts, and applying edits as you ask for them. Every chart creation or query change shows you a preview before anything is saved, and any change that would remove an existing connection between charts is flagged clearly before you confirm it. **AI Insights on every chart.** Each chart now includes an AI Insights card with four layers of context: * **Summary**: a concise description of what the chart shows. * **Business impact**: what the pattern in the data means for your business, including risks, opportunities, and suggested actions. * **How it's calculated**: how the insight is derived from your underlying data. * **Key highlights**: the most important patterns and changes to pay attention to. **Make dashboards match your brand.** Customize your primary color, a chart color palette of up to 8 colors, your dashboard background, and font pairing, and upload your organization's logo, with a live preview before anything is applied workspace-wide. **Jump straight to related data.** Configure Drill Through links so that clicking a data point takes you directly to a related dashboard or dataset, with the relevant filters (like a customer ID or date) carried over automatically. A single data point can have multiple Drill Through destinations to choose from, and you'll see a clear message if you don't have access to a configured target. This release also includes a range of accuracy and consistency improvements to how chart values are calculated and rendered, including better handling of null values, empty datasets, and edge cases when aggregating data. This release adds new ways to get a dashboard up and running fast, and smooths out onboarding for new users. **Turn a spreadsheet into a dashboard.** Upload a CSV or Excel file (including multi-sheet files) and the AI agent analyses it to automatically generate a relevant dashboard with appropriate charts, KPIs, and metrics. No data pipeline setup required. **Give Persona AI more context.** When generating a dashboard from an uploaded file, you can now also attach a reference document or Standard Operating Procedure. The agent uses it to better understand your terminology and business rules, resulting in more accurate, domain-aware dashboards. **One-click cube recommendations.** In the Knowledge Base, click Recommend Cubes to have AI analyze your available data sources and automatically generate relevant cube configurations, alongside the existing manual and natural-language creation options. **Simpler sign-up for new users.** New users can now register through a dedicated freemium sign-up flow and complete an onboarding form to personalize their workspace, all without needing an admin invitation. **More control over your charts.** * **Global filters**: add a filter that can control multiple charts at once, and choose exactly which charts respond to it. * **Table column filters**: filter table charts by individual columns for more granular data exploration. * **KPI chart drilldown**: configure a KPI chart to drill into more detail, either through specific drill members or by setting another chart or table as the drilldown target. This is the foundational release of the IdeaBoxAI platform, introducing AI-driven analytics, a smarter knowledge base, and conversational agents. **Agentic BI** Discontinued. AI-generated dashboards, built automatically from your business context and available data. * **AI-generated dashboards**: created automatically based on your business context and data. * **Automated metrics and visualizations**: the AI selects relevant KPIs, metrics, and chart types for you, with no manual configuration required. * **Manual and drag-and-drop editing**: fine-tune any dashboard by hand whenever you want full control. * **Foundational analytics and insights**: baseline trends, summaries, and performance indicators out of the box. * **AI assistant for dashboards**: an embedded assistant explains metrics, trends, anomalies, and relationships in plain language, so both business and technical users can work with the same dashboard. **A smarter Knowledge Base.** Structured, semantic analytics powered by a new Cube Modeler. * **Cube modeler creation**: build cube modelers through a simple, AI-powered workflow, right from the Knowledge Base. * **Natural language query-based creation**: describe what you need in plain language; no data-modeling expertise required. * **AI-powered cube generation**: an AI agent interprets your request, generates the cube configuration, and prepares it for deployment. * **Automatic deployment**: your cube modeler deploys automatically, with no manual setup. * **Seamless dashboard integration**: new cube-based metrics are instantly available for visualization in your Agentic BI dashboards. **Conversational AI agents.** A unified conversational and agent-driven intelligence layer across the platform. * **Natural language querying**: ask questions in plain language to explore your data, metrics, and insights. * **Insight explanation**: AI agents explain results, trends, and anomalies with context. * **Agent-orchestrated actions**: agents drive dashboard generation, cube modeling, and analytics workflows end-to-end. * **Context-aware intelligence**: agents draw on your Knowledge Base and cube models for accurate, consistent answers. **Platform foundations.** A core intelligence layer powering analytics, agents, and conversational workflows across the platform, a unified design system for a consistent experience everywhere, and an initial integration framework focused on onboarding your data. This release also includes early groundwork for an upcoming Agent Builder and Automations experience, single sign-on for all users, and a range of stability, reliability, and rendering improvements across the platform. # Sales Team Source: https://docs.ideaboxai.com/use-cases/actian-sales-copilot Learn how IdeaBoxAI's Sales Team Persona AI accelerates deal velocity across account research, deal health, technical discovery, and outbound prospecting, powered by 32 purpose-built skills. Every Account Executive knows the Sunday evening dread. Eight deals on the forecast. Two have gone quiet. One has a close date this Friday that everyone knows is slipping. Your manager wants a pipeline update at 9am Monday, and you still have not updated Salesforce from last week's calls. Every Sales Engineer knows the 4 pm scramble. A discovery call just ended. The prospect shared data volumes, query times, and stack details. You need to capture all of it, assess fit, and scope a POC before you forget what was said. Instead, you open a blank document and start typing from memory. Every SDR knows the Monday morning paralysis. Forty-seven leads in the queue. No signal on which ones are worth calling first. You pick one at random, spend 20 minutes researching, draft an email, and hope it lands. Multiply that by five and your morning is gone. The Sales Team Persona AI solves all three. It connects directly to Salesforce, Confluence, Google Drive, LinkedIn, and the public web. It understands your role, pulls live data, and answers your actual questions, in seconds, not hours. ## The Persona AI knows your role before you ask A Persona in IdeaBoxAI is the AI's understanding of who you are, what you need, and what your data looks like. When you log in as an Account Executive, you get a deal-management intelligence layer tuned to pipeline health, meeting prep, and forecast accuracy. When you log in as a Sales Engineer, you get a technical layer tuned to discovery, POC design, and objection handling. The Sales Team Persona AI ships with three purpose-built personas. Each connects to the data sources relevant to their role. The persona you are assigned to determines which data you see, which skills are available, and which scenario cards appear on your home screen. The following table summarizes the three personas. | **Persona** | **Focus** | **Key Needs** | | --------------------------- | ---------------------------- | ---------------------------------------------------------------------------------------------------- | | Account Executive (AE) | Deals, meetings, pipeline | Account research, deal health, meeting coaching, forecast prep, Salesforce hygiene | | Sales Engineer (SE) | Discovery, POC, objections | Discovery questions, technical fit assessment, POC scoping, competitive rebuttals, evidence assembly | | Sales Development Rep (SDR) | Leads, outbound, prospecting | ICP scoring, prospect profiling, outreach sequences, lead segmentation, Salesforce sync | Each persona also comes with pre-configured scenario cards on the home screen. Click any card to pre-fill the chat input with a common workflow prompt. You can also type your request directly in plain language; the Persona AI auto-detects the right skill. To browse all available skills, type `/` in the chat bar to open the skill picker. ## AE scenario: account research before a first meeting **Jordan Reeves ยท Account Executive** Mid-market territory ยท 14 active opportunities ยท First call with Meridian Health tomorrow at 10 am Jordan has a first meeting with Meridian Health in the morning. Normally, this means 45 minutes of tab-switching, LinkedIn for the org chart, the company website for strategic priorities, Salesforce for prior activity, and Google for recent news. By the time the research is done, there is no time left to build a talk track. **Jordan types:** *"Prep me for my meeting with Meridian Health tomorrow."* **The Persona AI responds:** > **Account brief, Meridian Health** > Prepared for Jordan Reeves ยท Meeting: Apr 24, 10:00 am ยท First call > > **Company snapshot** > > | **Attribute** | **Detail** | > | ------------------------- | ------------------------------------------------------------------- | > | Industry | Healthcare / Hospital systems | > | HQ | Minneapolis, MN | > | Revenue | \$2.1B (FY25) | > | Employees | 8,400 | > | Data infrastructure spend | Estimated \$14M annually (based on job postings and vendor signals) | > > **Strategic priorities and signals** > > * Announced a \$40M data modernization initiative; in Q1 earnings call, migrating from on-prem Teradata to a hybrid cloud architecture. > * Posted 6 data engineering roles in the last 90 days. Three mention "real-time analytics" and two mention "query performance at scale." > * New VP of Data Platform (Anika Patel) started 4 months ago. LinkedIn shows she came from UnitedHealth, where she led a similar migration. > > **Tech stack gaps** > > | **Current stack** | **Signal source** | **Actian fit** | > | ---------------------- | ------------------------------ | ----------------------------------------------------------------------------------- | > | Teradata (on-prem) | Job postings, Confluence notes | ๐ŸŸข High, migration path from Teradata is a proven use case. | > | Snowflake (evaluation) | Salesforce competitor field | ๐ŸŸก Moderate, they are evaluating, not committed. Real-time query gap is the angle. | > | Informatica ETL | Job postings | ๐ŸŸข High, DataConnect replacement opportunity if pipeline modernization is in scope. | > > **Competitive context** > Snowflake is listed as a competitor on this opportunity in Salesforce. Based on their data volumes (likely 500M+ rows given their patient record count) and their real-time requirements, the positioning angle is query latency at high concurrency, not migration ease. > > **Recommended talk track** > Lead with the data modernisation initiative, it is their stated priority, not something you are inferring. Acknowledge that Snowflake is strong for warehousing, then pivot to the real-time gap: "The question is not whether Snowflake is a good warehouse; it is whether it is the right engine for sub-second queries at your projected data volumes." > > Reference the FinServ customer running 3.8PB with sub-500ms average query time; healthcare and financial services share the same concurrency and compliance requirements. > > **Suggested questions for the call** > > * "What does your timeline look like for the Teradata migration, and what is driving the deadline?" > * "When you say real-time analytics, what query response time does your clinical team need to make decisions?" > * "How are you handling the compliance side of the migration? Is HIPAA driving any of your architecture choices?" > > **Suggested follow-ups:** Align my SE for this meeting ยท Build a pre-meeting checklist ยท Show me similar closed-won deals in healthcare ยท Save this brief to Salesforce Jordan reviews the brief in 4 minutes. The talk track is ready. The competitive angle is specific. The questions are tailored to what the prospect actually cares about, not a generic discovery script. **Jordan types:** *"I'm bringing my SE to this meeting. Brief my SE."* **The Persona AI responds:** > **SE alignment brief, Meridian Health** > For: SE assigned to Meridian Health ยท Meeting: Apr 24, 10:00 am ยท Stage 1 (First Call) > > **Account context** > > * Healthcare system, \$2.1B revenue, migrating from Teradata to hybrid cloud. > * Evaluating Snowflake. New VP of Data Platform (Anika Patel) is driving the initiative. > * Key signal: 6 data engineering hires in 90 days, 3 mention real-time analytics. > > **Your role in this meeting** > > * Listen for data volume numbers, current query times, and concurrency requirements. These are the inputs for the competitive angle. > * Be ready to address the Snowflake comparison, but do not lead with it. Wait for them to bring it up or for Jordan to set up the positioning. > > **Capabilities to demonstrate** > > * Vector AI DB real-time query performance at high row counts, the Teradata replacement story. > * DataConnect as an Informatica ETL alternative, only if they raise pipeline modernization. > > **POC design outline** > If the call goes well and they share specific data volumes, a 2-week POC on their patient analytics dataset would demonstrate the latency gap. Target metric: sub-2s query on a 30-day rolling window across their projected row count. > > **What NOT to lead with** > Do not lead with pricing or benchmarks in a Stage 1 call. Do not position against Snowflake unless they raise it first. Do not suggest a POC until discovery is complete. The SE alignment brief reuses research from the account brief generated in the same session. The Persona AI does not repeat the full data pull; it builds on what it already knows. The results: * Account research reduced from 45 minutes to under 5 minutes. * Talk track tailored to the prospect's stated priorities, competitive situation, and tech stack, not a generic pitch. * SE briefed with role-specific context and a clear POC outline in 30 seconds. * All research auto-saved to Salesforce and Google Drive. ## AE scenario: deal health check before forecast **Rachel Torres ยท Account Executive** Enterprise territory ยท 11 active opportunities ยท Forecast call with VP at 2 pm Rachel has a forecast call in two hours. She knows her pipeline has problems but cannot pinpoint which deals are actually at risk versus which ones just feel slow. Before IdeaBoxAI, she would open each opportunity in Salesforce one by one and try to piece together the story from activity logs and her own memory. **Rachel types:** *"Run a deal health check on NovaTech. They've gone quiet, and I'm worried about the close date."* **The Persona AI responds:** > **Deal health scorecard, NovaTech** > Opportunity: NovaTech, Enterprise Data Platform ยท ARR: \$185,000 ยท Stage 4 ยท Close date: May 16 > > **Overall score: 54 / 100, Recovering** ๐ŸŸก > > | **Dimension** | **Score** | **Evidence** | > | --------------------- | ----------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | > | Activity momentum | 35 / 100 ๐Ÿ”ด | Last activity was 18 days ago (email from Rachel, no reply). Prior to that, 3 calls in 2 weeks. The drop is sharp. | > | Stakeholder coverage | 45 / 100 ๐Ÿ”ด | Only 1 contact engaged (David Cho, Sr. Data Engineer). No executive sponsor identified. No procurement contact. Single-threaded. | > | Artifact completeness | 70 / 100 ๐ŸŸก | Account brief on file. Discovery notes on file. Missing: mutual action plan, proposal, executive sponsor alignment. | > | Close date integrity | 50 / 100 ๐ŸŸก | Close date has been pushed once already (from Apr 30 to May 16). At current velocity, May 16 requires a stage progression within 5 business days, which has not happened. | > | Competitive exposure | 65 / 100 ๐ŸŸก | Databricks listed as competitor. No update on competitive status since Stage 2. Last mention: "They're also talking to Databricks but haven't started a POC." That was 4 weeks ago. | > > **Risk flags** > > * **Stale activity (critical):** 18 days since last response. Your average deal at Stage 4 has contact every 4.2 days. NovaTech is 4x the norm. > * **Single-threaded (high):** Only David Cho is engaged. He is a Sr. Data Engineer, not a decision-maker. No executive sponsor on record. > * **Close date drift (medium):** Already pushed once. Current trajectory does not support May 16 without immediate re-engagement. > > **Benchmark:** Deals at the same stage and ARR that closed-won had an average health score of 78. NovaTech is 24 points below that benchmark. > > **Suggested follow-ups:** View next actions for this deal ยท Draft a re-engagement email for David Cho ยท Generate a mutual action plan ยท Update Salesforce with forecast category Rachel clicks the suggested follow-up. **Rachel types:** *"What should I do today to save this deal?"* **The Persona AI responds:** > **Next actions, NovaTech** > Based on deal health score (54/100) and 3 open risk flags. > > **Today**
# Action Channel Why Exact words
1 Re-engage David Cho Call (not email) 18 days silent. Email has failed; David has not replied to the last 2. A call breaks the pattern. "David, I wanted to check in directly. Last time we spoke, you were comparing timelines with your Databricks evaluation. I have some new benchmark data that might help with that. Do you have 10 minutes this week?"
2 Ask David for an exec intro Same call Single-threaded risk. You need a second contact before this deal can progress. "One thing that would help me build the best proposal for your team: would it make sense to include your VP in a 15-minute alignment call? I want to make sure the business case lands, not just the technical one."
3 Update Salesforce Persona AI Log today's call attempt, update next step date, flag the single-threaded risk in the opportunity notes. I can update Salesforce for you after the call; just tell me what happened.
> **This week**
# Action Goal
4 Generate a mutual action plan working back from May 16. Creates a shared timeline that forces both sides to commit to milestones, or reveals that May 16 is not realistic.
5 Draft a re-engagement email with the Databricks benchmark as the hook. Backup if the call does not connect. Subject references the benchmark, not "checking in."
6 Update forecast category to Best Case with a note. Score is 54 with active re-engagement. Commit requires a score above 90 with no open flags.
Rachel walks into her forecast call with a clear story: NovaTech is at risk, here is why, here is what she is doing about it today, and here is what the forecast category should be, with evidence behind each point. The results: * Deal risk surfaced across 5 dimensions with specific Salesforce evidence, not gut feel. * 18-day silence flagged and benchmarked against similar deals at the same stage. * Prioritized action plan split into today and this week, with exact words for each conversation. * Forecast category recommended with rationale the VP can audit. ## SE scenario: post-discovery debrief and POC scoping **Marcus Liu ยท Sales Engineer** Assigned to 6 active evaluations ยท Just finished a discovery call with DataStream Analytics Marcus just got off a 45-minute technical discovery call with DataStream Analytics. The champion shared volumes, query requirements, stack details, and a decision timeline. Marcus needs to capture all of it before the details blur, and assess whether this is a strong fit. **Marcus types:** *"Call done, here's what I learned. DataStream Analytics, VP of Engineering is the champion. They're running Redshift today with about 800M rows, growing 15% quarterly. Current dashboard queries take 12-18 seconds; they need sub-2 seconds. They're evaluating us alongside BigQuery. Decision by end of June, POC needs to start by mid-May. Stack is Python, Airflow, dbt. Main concern is migration complexity from Redshift."* **The Persona AI responds:** > **Post-discovery summary, DataStream Analytics** > > **Key facts card** > > | **Attribute** | **Detail** | > | ------------------- | --------------------------------- | > | Company | DataStream Analytics | > | Champion | VP of Engineering | > | Current stack | Redshift, Python, Airflow, dbt | > | Data volume | 800M rows, growing 15% quarterly | > | Current query time | 12โ€“18 seconds (dashboard queries) | > | Required query time | Sub-2 seconds | > | Competitor | BigQuery | > | Decision deadline | End of June | > | POC start window | Mid-May | > > **Actian fit assessment: Strong** ๐ŸŸข > > The fit is strong for three reasons: > > * Their query performance gap is significant. 12โ€“18 seconds down to sub-2 seconds is exactly the use case where Vector AI DB outperforms both Redshift and BigQuery at their row count. > * 800M rows growing 15% quarterly puts them at 1.2B rows within 12 months. BigQuery charges per query at that scale. Vector AI DB has no per-query cost; the TCO argument strengthens as they grow. > * Their stack (Python, Airflow, dbt) integrates cleanly. No exotic dependencies. > > **Migration risk note:** The champion flagged migration complexity as their main concern. This is a category 2 objection (Migration Risk). Acknowledge the real cost first; do not minimize it. Then reframe: the question is not whether migration has a cost, but whether staying on Redshift at 12โ€“18 second query times has a higher one. > > **Recommended next step:** Generate a POC proposal scoped to their dashboard use case. Target metric: sub-2s query on a 30-day rolling window at 800M rows. > > **Suggested follow-ups:** Generate a POC proposal for DataStream ยท Draft an email to the champion with POC scope ยท Update Salesforce stage ยท Pull benchmark data for Redshift vs Vector AI DB Marcus clicks the POC suggestion. **Marcus types:** *"Generate a POC proposal for DataStream."* **The Persona AI responds:** > **POC scope card, DataStream Analytics** > > | **Attribute** | **Detail** | > | --------------------- | --------------------------------------------------------------------------------------------- | > | Dataset | Production dashboard dataset, live Redshift export (preferred over synthetic). | > | Dataset size | 800M rows current, test at 1.2B to demonstrate growth headroom. | > | Success metric | Sub-2s query on 30-day rolling window aggregation at 800M rows. Stretch: sub-1s at 1.2B rows. | > | Champion | VP of Engineering | > | Duration | 2 weeks (standard, integration complexity is low given Python/Airflow/dbt stack). | > | DataConnect | Not included. Their Airflow pipeline handles ingestion today; no ETL replacement needed. | > | Vector AI DB use case | Real-time dashboard analytics replacing Redshift. | > > **Suggested next steps** > > 1. Send the POC scope to the champion for sign-off on dataset and success metric. > 2. Schedule a 30-minute POC kickoff call with Marcus and the champion's engineering lead. > 3. Request a Redshift export of 30 days of dashboard query data for baseline comparison. > 4. Draft a post-call email to the champion confirming the POC timeline and what both sides need to deliver. The POC proposal pulls directly from the discovery debrief; dataset size, query requirements, and stack details are carried forward automatically. The success metric is quantified, not qualitative. The results: * Discovery findings captured and structured in under 60 seconds, no manual note formatting. * Actian fit scored with specific reasoning tied to data points from the call, not generic categories. * POC scoped with a quantified success metric, realistic timeline, and clear next steps. * All findings auto-logged to Salesforce. ## AE scenario: post-meeting log and follow-up email **Sarah Park ยท Account Executive** Enterprise territory ยท Just finished a Stage 3 call with Vertex Financial Sarah just walked out of a meeting with Vertex Financial. She has 12 minutes before her next call and needs to log what happened, update Salesforce, and send a follow-up email. Before IdeaBoxAI, two of those three would not get done until the end of the day, if at all. **Sarah types:** *"Meeting done with Vertex Financial, here's what happened. Good energy overall. CFO joined for the first 15 minutes, which was unexpected. She asked about SOC 2 compliance and whether we support data residency in the EU. I confirmed SOC 2 Type II and said I'd follow up on EU residency. David (champion) confirmed they want to move to POC but need budget approval from the CFO first. Competitor is Snowflake; they've done a POC with Snowflake already but weren't happy with query latency on their risk modeling workload. Next step is a technical deep-dive with their data engineering team next week. One risk: David mentioned the project could get deprioritized if Q2 revenue comes in below target."* **The Persona AI responds:** > **Post-meeting log, Vertex Financial** > Meeting date: Apr 23, 2026 ยท Stage 3 ยท Attendees: Sarah Park (AE), David (Champion), CFO (new contact) > > **What went well** > > * CFO joined unprompted, signals executive interest. This is unusual at Stage 3 and suggests the initiative has visibility at the C-level. > * Champion confirmed intent to move to POC. Budget approval is the remaining gate. > * Snowflake POC completed but did not meet latency requirements on risk modelling. This is a significant competitive advantage; they have already experienced the gap you are positioning against. > > **Deal status** > Stage stays at 3. Move to Stage 4 when budget approval is confirmed, and the technical deep-dive is scheduled with a date. > > **Open risks** > > | **Risk** | **Severity** | **Specific action** | **Exact words** | > | ------------------------------------- | ------------ | ----------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------- | > | EU data residency question unanswered | ๐ŸŸก Medium | Follow up within 24 hours. Delayed response on a CFO question erodes trust. | Include in the follow-up email with a clear answer or a specific date when you will have one. | > | Budget deprioritization if Q2 misses | ๐ŸŸก Medium | Accelerate the timeline. The longer this deal sits at Stage 3, the higher the risk of a freeze. | In your next call with David: "What does the budget approval process look like, and is there a way to get sign-off before the Q2 review?" | > | CFO is a new stakeholder, no record | ๐ŸŸข Low | Add the CFO as a contact in Salesforce today. Log her SOC 2 and residency questions. | I can add her to the opportunity record when you confirm the update. | > > **Immediate next actions** > > 1. Send follow-up email to David confirming the technical deep-dive and addressing the EU residency question. (Due: today.) > 2. Add CFO as a contact in Salesforce with meeting notes. (Due: today.) > 3. Schedule the technical deep-dive with David's data engineering team. (Due: this week.) > > **Suggested follow-ups:** Draft the follow-up email ยท Update Salesforce with these notes ยท Generate a mutual action plan for Vertex ยท Prepare objection prep for the next call Sarah immediately follows up. **Sarah types:** *"Draft the follow-up email to David."* **The Persona AI responds:** > **Email draft, Post-meeting follow-up** > > **To:** David ยท Vertex Financial > **Subject:** Vertex risk modeling POC, next steps and residency answer > > David, > > Good speaking with you and your CFO today. Two follow-ups from the conversation. > > First, on EU data residency, I am confirming the specifics with our compliance team and will have a documented answer for you by Friday. SOC 2 Type II certification is current, and I can share the report directly if helpful for your CFO's review. > > Second, the technical deep-dive. Are you available next Tuesday or Wednesday for a 60-minute session with your data engineering team? Marcus (our SE) will walk through the architecture and how it maps to your risk modeling workload, specifically the latency gap you experienced in the Snowflake POC. > > Let me know what works. > > Sarah The Persona AI requires explicit confirmation before writing anything to Salesforce. It shows you exactly what will be logged: activities, notes, stage changes, and forecast updates, and waits for your approval. The results: * Meeting logged with structured outcomes, risk flags, and next actions in under 90 seconds. * Follow-up email drafted with a specific hook (the Snowflake latency gap) and a concrete ask (deep-dive date), no generic opener. * CFO flagged as a new stakeholder with her specific concerns captured for future reference. * 12-minute window between meetings used productively instead of lost. ## SDR scenario: weekly lead scoring and outbound **Priya Mehta ยท Sales Development Rep** Assigned territory: West Coast mid-market ยท 47 unworked leads in Salesforce Priya starts every Monday the same way, staring at a list of leads with no idea which ones to call first. She used to pick accounts alphabetically or by the order they came in. Half her outreach went to companies that were never going to buy. The other half went out with a generic template. **Priya types:** *"Set up my week. Where should I start?"* **The Persona AI responds:** > **Weekly priority checklist, Priya Mehta** > Territory: West Coast mid-market ยท 47 assigned leads ยท Week of Apr 21 > > **Step 1 of 10: ICP scoring complete** > 47 leads scored across 4 dimensions (Firmographic, Technographic, Intent Signals, Win Pattern Match). > > | **Tier** | **Count** | **Action** | > | --------------- | --------- | ------------------------------------------------------------- | > | ๐Ÿ”ฅ HOT (75+) | 5 | Prioritize above everything else this week. | > | ๐ŸŸก WARM (50โ€“74) | 12 | Reach out this week if time permits, next week at the latest. | > | โ„๏ธ LOW (\< 50) | 30 | Route to segmentation, do not discard without guidance. | > > **Top 5 HOT accounts** > > | **Account** | **Score** | **Top signal** | **Recommended first action** | > | ----------------------- | --------- | -------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------ | > | Cascade Data Systems | 92 | Posted VP of Data Platform role 3 weeks ago. Stack matches (Teradata, Python). Closest closed-won match: MedTech Corp (\$142K ARR, 4-month cycle). | Build a prospect profile and draft a personalized email to the hiring manager. | > | Ridgeline Analytics | 88 | Raised \$28M Series C 6 weeks ago. Job postings mention "real-time analytics at scale." | Research the VP of Engineering on LinkedIn, then build an outreach sequence. | > | Apex Financial Group | 84 | Existing Snowflake customer posting about query performance issues on their engineering blog. | Draft a cold email using the blog post as the hook. | > | Pacific Logistics | 79 | Data engineering team doubled in 6 months. Current stack includes Redshift. | Build a prospect profile to identify the right contact. | > | Westfield Manufacturing | 76 | SAP integration need identified from job postings. Manufacturing vertical, 38% historical close rate. | Research the Head of Data and draft outreach leading with SAP integration. | > > **Remaining 42 leads:** 12 WARM accounts summarised with top signal. 30 LOW accounts ready for segmentation when you finish Tier 1 outreach. > > **Suggested follow-ups:** Show me the full score for Cascade Data Systems ยท Build a prospect profile for the VP of Data Platform at Cascade ยท Start outreach for the top 5 ยท Segment my low-fit leads Priya starts with her top lead. **Priya types:** *"Build a prospect profile for the VP of Data Platform at Cascade Data Systems."* **The Persona AI responds:** > **Prospect profile, Cascade Data Systems** > > **Company snapshot** > > | **Attribute** | **Detail** | > | ------------- | ------------------------------------------ | > | Industry | Data infrastructure / SaaS | > | HQ | Portland, OR | > | Revenue | \$85M (estimated) | > | Employees | 420 | > | Data stack | Teradata (migrating), Python, Apache Spark | > > **Trigger events (last 90 days)** > > * Posted VP of Data Platform role 3 weeks ago, job description mentions "migrate from legacy warehouse to a modern, high-performance analytics engine." > * Hired 4 data engineers in the last 60 days. > * CTO spoke at a Portland data meetup about "the cost of slow queries on product decisions." > > **Tech stack signals** > > | **Technology** | **Signal** | **Actian fit** | > | ----------------- | ------------------------------------------- | ---------------------------------------------------------------- | > | Teradata (legacy) | Job postings explicitly reference migration | ๐ŸŸข HIGH, proven migration path, DataConnect accelerates ETL lift | > | Apache Spark | Engineering blog | ๐ŸŸก MODERATE, Vector AI DB complements Spark as the query layer | > | Python | Job postings | ๐ŸŸข HIGH, native Python SDK, clean integration | > > **Contact profiling** > > * **Target:** VP of Data Platform (role posted, not yet filled, hiring manager is likely the CTO or a Sr. Director of Engineering). > * **Pain hypothesis:** Migrating from Teradata means they need a replacement that delivers query performance at scale without the licensing cost. The CTO's public comments about "slow queries" confirm this is a business-level pain, not just a technical wish. > > **Personalisation hooks** > > 1. **Trigger:** The VP of Data Platform job posting mentions "high-performance analytics engine", this is their own language for what Vector AI DB delivers. > 2. **Pain:** Their CTO publicly called out slow queries as a product blocker. Query latency is not an infrastructure issue for them; it is a revenue issue. > 3. **Proof:** MedTech Corp (similar size, similar Teradata migration) closed at \$142K ARR in 4 months. Sub-500ms query times on 600M rows. > > **Suggested follow-ups:** Draft an outreach email using the CTO's meetup talk as the hook ยท Build a LinkedIn message for the CTO ยท Create a full outreach sequence ยท Save this profile to Salesforce Priya drafts the email in one click, sends the LinkedIn message, and loads the full sequence into Salesforce, all before 9:30 am. The rest of her top 5 follow the same pattern. The results: * 47 leads scored and ranked in under 8 seconds, with the specific signal driving each score. * Top 5 accounts identified with a clear first action for each, grounded in real signals. * Prospect profile built from Salesforce, LinkedIn, and the public web in under 30 seconds, including personalization hooks ready to paste into outreach. * Monday morning paralysis replaced with a structured 10-step plan. ## Skills that power all three personas Each persona in IdeaBoxAI is powered by a set of Skills, AI capabilities that define how the Persona AI handles specific types of requests. The Sales Team Persona AI ships with 32 skills out of the box, split across the three personas. You can trigger a skill in three ways: * Click a **scenario card** on the home screen to pre-fill the prompt and run the skill immediately. * Type your request in plain language. The Persona AI auto-detects the appropriate skill from your message. * Type `/` in the chat bar to open the skill picker and select a skill by name. You do not need to memorize skill names or commands. Type what you need in plain language, and the Persona AI matches your intent to the right skill automatically. ### Account Executive skills The following skills handle deal management, meeting preparation, and pipeline workflows. | **Skill** | **What it does** | | ------------------------ | ------------------------------------------------------------------------------------------------------------------------------------- | | Account Brief Generation | Pulls signals from Salesforce, LinkedIn, job postings, and the web to build a 6-section account brief in under 5 minutes. | | SE Alignment Brief | Creates a pre-meeting brief for the Sales Engineer, account context, capabilities to demo, POC outline, and positioning notes. | | Meeting Coaching | Returns the single most important outcome for an upcoming meeting based on deal stage, with exact words to close on it. | | Objection Prep | Generates a prioritized list of likely objections with probability, underlying concern, and conversational responses. | | Pre-Meeting Checklist | Produces a 3-section checklist, materials to bring, Salesforce hygiene items, and talking points to memorize. | | Post-Meeting Log | Turns a freeform meeting debrief into a structured outcome log, deal status, risk flags, and next actions, auto-logged to Salesforce. | | Email Draft | Drafts any sales email, cold outreach, warm follow-up, post-meeting, or post-discovery, under 150 words with a signal-based hook. | | Deal Scorecard | Scores deal health across 5 dimensions (activity, stakeholders, artifacts, close date, competition) with evidence from Salesforce. | | Deal Next Actions | Generates a prioritized action list split into today and this week, with channel, exact words, and the goal each action achieves. | | Re-engage Email | Drafts a direct re-engagement email for a prospect who has gone quiet, no apology, clear reason to reply. | | Mutual Action Plan | Builds a 5-milestone close plan working backward from the target close date, with owners and deliverables for both sides. | | Salesforce Update | Logs session activity to Salesforce, meeting notes, calls, stage changes, and forecast category. Requires explicit confirmation. | | Forecast Guidance | Recommends a forecast category (Commit, Best Case, Pipeline) with evidence-backed talking points for the pipeline review. | | Call Script | Writes a branching call script with a direct opening, conditional paths, a named key ask, and a committed next step with a date. | | Deal Artifact Checklist | Audits which deal documents exist, which are missing, and which gaps are blocking stage progression. | | PDF Export | Exports any brief, scorecard, or plan as a formatted, single-page PDF with branding and a confidential watermark. | ### Sales Engineer skills The following skills handle technical discovery, POC design, and competitive objection handling. | **Skill** | **What it does** | | ----------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------- | | Discovery Question Set | Generates 15โ€“25 tailored discovery questions adapted to industry, tech stack, and deal stage, each annotated with why to ask it. | | Discovery Note Template | Creates a live note-taking template with questions as headers and blank answer fields, auto-saved to Salesforce on completion. | | Post-Discovery Summary | Extracts technical facts from a freeform call debrief, scores Actian fit, and generates a POC recommendation automatically. | | POC Recommendation | Designs a POC scope, dataset, success metric, duration, champion, and products, built to demonstrate advantage against evaluation criteria. | | Rebuttal Card | Returns a structured response to any technical objection in under 3 seconds, including per-competitor positioning for Snowflake, Databricks, and BigQuery. | | Evidence Assembly | Retrieves specific benchmarks, case studies, and certifications matched to the objection category and the prospect's industry. | | Objection Email | Drafts a post-call follow-up addressing technical objections raised during a session, with supporting evidence as named attachments. | | Objection Tracker | Tracks all open objections across a deal, who raised them, what was shared, and whether each one is resolved. | ### Sales Development Rep skills The following skills handle lead scoring, prospect research, and outbound sequencing. | **Skill** | **What it does** | | ------------------------- | ------------------------------------------------------------------------------------------------------------------------------- | | Prospect Profile | Builds an enriched profile, company snapshot, trigger events, tech stack signals, pain hypothesis, and 3 personalization hooks. | | LinkedIn Message | Writes a 3-sentence LinkedIn message with a signal-based hook, a value sentence, and a low-friction question as the CTA. | | Outreach Sequence | Builds a 5-touch, 14-day outreach sequence across email, LinkedIn, and phone, each touch anchored to a real signal. | | ICP Score (Batch) | Scores an entire lead list against the ICP model and ranks every account by conversion likelihood with heat labels. | | ICP Score (Single) | Gives a full ICP breakdown on one account, every dimension scored, with the closest closed-won match named as reference. | | Low-Fit Segmentation | Segments low-scoring leads into Nurture, Disqualify, and Needs More Data, with specific next steps for each group. | | Weekly Priority Checklist | Sets up the week with a 10-step prioritization routine, from scoring leads to drafting emails to loading sequences. | | Salesforce Lead Sync | Saves outbound research, email drafts, trigger events, and ICP scores to Salesforce. Requires explicit confirmation. | ## Connected to your sales data The Sales Team Persona AI is grounded in your live data, not a sample, not a simulation. It connects to the systems your sales team already uses, pulling information in real time to inform every response. When an AE asks "prep me for my meeting with Meridian Health," the Persona AI queries Salesforce for opportunity data, searches LinkedIn for recent activity, scans job postings for tech stack signals, and pulls prior notes from Confluence, all in parallel, in under 5 minutes. Each persona only sees the data relevant to their role. An AE sees opportunities, contacts, and deal activity. An SE sees technical notes, discovery records, and POC status. An SDR sees leads, ICP scores, and outreach history. The platform connects through the following layers: * Salesforce for CRM data, opportunities, contacts, leads, activity history, and forecast categories. * Confluence for internal knowledge, battlecards, case studies, prior SE notes, and RFP templates. * Google Drive for deal documents, proposals, account briefs, and mutual action plans. * LinkedIn for prospect research, role changes, company signals, and professional context. * Public web for company intelligence, news, job postings, financial filings, and tech stack signals. ## Getting started Setting up the Sales Team Persona AI for your team takes less than a day. Link Salesforce, Confluence, Google Drive, and LinkedIn from the Connections settings in the Admin Console. Each integration is configured per-organization and tested to confirm data visibility. Set up Account Executive, Sales Engineer, and SDR personas in the Admin Console. Assign the relevant skills and knowledge base to each. The knowledge base scopes each persona to their relevant data sources automatically. Click the knowledge base generation button. The platform maps your Salesforce object structure, indexes your Confluence content, and generates the cubes needed for Agent BI dashboards. This takes under 10 minutes. Add your sales team members and assign them to their persona. They log in and see their role-specific Persona AI immediately. Scenario cards guide them through the most common workflows from day one. Start on the IdeaBoxAI dev environment to validate connections and test persona responses against your real Salesforce data. Once the team is confident, promote to production. The Sales Team Persona AI is available now as part of IdeaBoxAI's Persona AI suite. Contact the IdeaBoxAI team to set up a live demo connected to your Salesforce environment. ## Next steps Explore these guides to learn more about the platform capabilities behind the Sales Team Persona AI. Create and configure role-specific personas in the Admin Console. Assign skills to personas and customize how the Persona AI handles requests. Connect structured data sources and generate knowledge bases for grounded responses. Understand the measurable outcomes the Conversational AI delivers. # Automated reporting Source: https://docs.ideaboxai.com/use-cases/automated-reporting Learn how teams use IdeaBoxAI Automations to generate and distribute reports without manual effort. ## Overview This use case demonstrates how a multi-location retail chain used IdeaBoxAI Automations to eliminate 15+ hours per week of manual report generation, delivering daily performance reports, weekly inventory summaries, and monthly executive dashboards automatically to 40+ stakeholders โ€” improving data freshness from 48-hour delays to real-time delivery. ## Background **Evergreen Retail Group**, a regional retail chain operating 28 stores across the Southeast, relied heavily on operational and performance reports to drive decision-making. Store managers needed daily sales summaries to adjust staffing and inventory. Regional directors required weekly performance comparisons across locations. The executive team expected monthly board-ready reports with revenue trends, margin analysis, and year-over-year comparisons. Before implementing IdeaBoxAI Automations, all of these reports were generated manually by two business analysts who would: 1. Export data from the POS system (PostgreSQL), inventory management system (MySQL), and HR platform (CSV exports) 2. Build Excel pivot tables, charts, and formatted summaries 3. Copy-paste data into PowerPoint templates for executive presentations 4. Email individual reports to 40+ stakeholders based on role and location 5. Field follow-up questions and requests for custom date ranges or filtered views This manual process consumed 15โ€“18 hours per week, introduced frequent copy-paste errors, and created significant delays โ€” by the time regional managers received their reports, the data was already 24โ€“48 hours old, limiting their ability to respond quickly to trends or issues. ## The challenge Evergreen Retail Group faced four core reporting challenges: 1. **Time-intensive manual work** โ€” The two analysts spent nearly half their workweek building reports, leaving little time for actual data analysis or strategic projects. During busy periods (e.g., Black Friday, holiday season), report generation became a bottleneck that delayed decision-making. 2. **Data staleness** โ€” Reports were generated once per day (morning batch run) or once per week (Monday morning). If a store experienced an unexpected sales spike or inventory shortage on Tuesday afternoon, managers wouldn't see the issue reflected in their reports until Wednesday morning at the earliest. 3. **Inconsistent formatting and errors** โ€” Manual copy-paste workflows introduced frequent errors: transposed numbers, incorrect date ranges, mismatched chart labels, and inconsistent formatting across reports. These errors eroded trust and required additional time to verify and correct. 4. **Limited personalization** โ€” Every store manager received the same report template, even though each location had different priorities (e.g., stores in college towns cared about student traffic patterns, coastal stores tracked tourism seasonality). Custom reports required special one-off requests that the analysts couldn't accommodate at scale. ## The solution Evergreen Retail Group implemented IdeaBoxAI Automations to build a fully automated reporting pipeline that generated, personalized, and distributed reports to the right stakeholders at the right time โ€” with zero manual intervention. ### Implementation approach The team created three Knowledge Bases in IdeaBoxAI: * **Sales & POS KB** โ€” Connected to their PostgreSQL POS database with transaction-level sales data, payment methods, refunds, and customer counts. * **Inventory KB** โ€” Connected to their MySQL inventory system tracking SKU-level stock, reorder points, supplier lead times, and transfer history. * **Staffing KB** โ€” Built from weekly CSV exports containing employee schedules, labor hours, and payroll costs by store location. These Knowledge Bases provided real-time access to operational data without requiring the automation to write custom SQL queries. Using Agentic BI, the team created three master dashboard templates: * **Daily Store Performance** โ€” Revenue, transactions, average ticket, top SKUs, and hourly sales patterns. * **Weekly Regional Summary** โ€” Store-by-store comparison of revenue, margin, inventory turnover, and labor costs. * **Monthly Executive Report** โ€” High-level KPIs, trend analysis, year-over-year growth, and AI-generated insights. Each dashboard was designed with filters, drill-down capabilities, and clean formatting suitable for both web viewing and PDF export. The team used IdeaBoxAI Automations to schedule report generation and distribution: * **Daily Store Reports** โ€” Triggered every morning at 6:00 AM, querying sales data from the previous day. Each store manager received a personalized PDF report via email showing only their location's data. * **Weekly Regional Reports** โ€” Triggered every Monday at 7:00 AM, aggregating the previous week's performance. Regional directors received a single report comparing all stores in their territory. * **Monthly Executive Reports** โ€” Triggered on the 1st of each month at 8:00 AM, generating a board-ready PowerPoint-style summary with charts, tables, and AI-generated commentary. Delivered to the executive team and board members. Each automation included conditional logic: if revenue dropped more than 15% week-over-week, the report would highlight the variance in red and trigger an alert to the regional director. In addition to scheduled email delivery, the team deployed a self-service reporting portal where any stakeholder could: * View the latest version of their dashboard in real time (no waiting for the next scheduled run) * Apply custom filters (e.g., *Show me last 30 days* or *Compare this week to the same week last year*) * Download reports as PDF or Excel on demand * Subscribe to additional report types or change their delivery preferences This eliminated the need for analysts to field ad-hoc report requests. ### Key configurations Evergreen Retail Group leveraged several IdeaBoxAI Automation features to optimize their reporting workflow: * **Personalized report distribution** โ€” Each automation dynamically filtered data by store location or region before generating the report. Store managers only saw their own store's data, while regional directors received aggregated multi-store summaries. This personalization was configured using dynamic filters based on the recipient's role and assigned locations. * **Conditional alerting** โ€” Reports included smart highlighting: metrics that exceeded targets were shown in green, underperforming metrics in red. If key thresholds were breached (e.g., inventory below reorder point, labor costs exceeding budget), the automation sent an immediate Slack notification to the relevant manager in addition to the scheduled email. * **AI-generated commentary** โ€” Each executive report included a natural language summary generated by IdeaBoxAI's AI Assistant, such as: *"Revenue increased 8.2% compared to last month, driven primarily by strong performance in the Electronics and Home Goods categories. Store #14 (Savannah location) showed the highest growth at +18%, while Store #7 (Athens location) declined -5%, likely due to scheduled renovations."* * **Version history and audit trail** โ€” Every generated report was automatically archived with a timestamp, enabling the team to review historical reports, compare trends over time, or investigate discrepancies. ## Results Within 90 days of deploying automated reporting, Evergreen Retail Group achieved measurable efficiency gains and improved decision-making speed: | Metric | Before | After | Improvement | | ----------------------------------- | --------- | --------- | ----------- | | **Weekly hours spent on reports** | 15โ€“18 hrs | 2 hrs | -89% | | **Report delivery delay** | 24โ€“48 hrs | Real-time | -100% | | **Copy-paste errors per month** | 12โ€“15 | 0 | -100% | | **Stakeholder satisfaction (CSAT)** | 3.1/5 | 4.7/5 | +52% | | **Ad-hoc report requests** | 45/month | 6/month | -87% | | **Time to respond to issues** | 36 hrs | 4 hrs | -89% | ### Qualitative outcomes Beyond the quantitative metrics, the team and stakeholders reported several qualitative benefits: * **Analyst capacity freed for strategic work** โ€” The two analysts who previously spent half their time building reports now focused on exploratory analysis, forecasting models, and business recommendations โ€” significantly increasing their impact and job satisfaction. * **Faster issue detection** โ€” Store managers noticed inventory shortages, staffing gaps, and sales anomalies within hours instead of days, enabling proactive corrective action before issues escalated into customer complaints or lost revenue. * **Consistent and trustworthy data** โ€” Automated report generation eliminated copy-paste errors and ensured that all stakeholders viewed the same source of truth, reducing confusion and disputes during meetings. * **Empowered self-service culture** โ€” Department heads and store managers became comfortable accessing the self-service portal to explore data on their own, reducing their dependency on the analytics team and fostering a more data-driven culture across the organization. ## Key takeaways This use case highlights three core capabilities of IdeaBoxAI Automations that are especially valuable in reporting contexts: 1. **Elimination of manual toil** โ€” By automating repetitive report generation, formatting, and distribution, Evergreen freed nearly 16 hours per week of analyst time โ€” time that could be redirected toward higher-value analysis and strategic initiatives. 2. **Real-time data delivery** โ€” Scheduled automations ensured that stakeholders received fresh data at the optimal time (e.g., daily reports delivered first thing in the morning, weekly reports on Mondays before planning meetings), enabling faster, more informed decision-making. 3. **Personalization at scale** โ€” Dynamic filtering and conditional logic allowed a single automation workflow to serve 40+ stakeholders with individualized, role-appropriate reports โ€” something that would be impossible to maintain manually. ## Technical considerations For teams evaluating automated reporting implementations, Evergreen's experience surfaced several important technical lessons: * **Start with existing templates** โ€” Rather than redesigning all reports from scratch, the team started by replicating their existing Excel and PowerPoint formats in Agentic BI. This reduced change management friction and made adoption easier. * **Test with historical data first** โ€” Before scheduling automations to run in production, the team tested each workflow by running it against historical date ranges and comparing the output to manually generated reports. This validation step caught several edge cases (e.g., handling stores that were temporarily closed, accounting for DST time shifts). * **Build in redundancy for critical reports** โ€” For high-stakes reports (e.g., board presentations, investor updates), the team configured dual-delivery: the automation emailed the report and also uploaded it to a shared drive as a backup, ensuring no single point of failure. * **Iterate based on feedback** โ€” After launch, the team held weekly feedback sessions with stakeholders to identify missing metrics, confusing layouts, or desired customizations. They iterated on the dashboards and automation logic over 6 weeks before considering the system "production-ready." ## Related resources Learn how to create your first automation workflow in IdeaBoxAI. Step-by-step tutorial for creating and configuring your first automation. Create dashboard templates that can be automated and distributed via email or portal. Configure filters for personalized, role-based dashboard views. # Driving business outcomes Source: https://docs.ideaboxai.com/use-cases/business-outcome Understand the business problems the Persona AI solves and the measurable outcomes it delivers across sales teams. The Persona AI is designed to address specific, measurable inefficiencies across sales workflows. This page maps each problem area to the platform capabilities that solve it. ## Problems addressed The following table maps each problem area to its current state and the business impact it creates. | Problem Area | Current State | Business Impact | | ------------------------------ | --------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------ | | **Manual research overhead** | AEs spend 30-60 min per account on research with no standard format. | Inconsistent discovery, missed signals, lost selling time. | | **Invisible deal risk** | No systematic tracking of missing artefacts, stale activity, or single-threaded deals. | Surprise slips, forecast inaccuracy, avoidable losses. | | **Generic outreach and demos** | SDRs reuse stale templates; SEs configure demos manually without deal context. | Low reply rates, poor demo-to-POC conversion. | | **Fragmented knowledge** | Battlecards, RFP answers, and objection responses scattered across multiple systems and tribal knowledge. | Inconsistent positioning, slow responses, long SE ramp time. | ## How the Persona AI solves each problem * **Research in under 5 minutes**: The Account Research skill generates structured value briefs, company snapshot, strategic priorities, tech stack, pain-to-product mapping, and talk tracks, in under 5 minutes instead of 30-60 minutes. * **Proactive deal risk detection**: The Deal Health Scorecard flags missing artifacts, stale activity, single-threaded contacts, and incomplete qualification fields. Momentum alerts fire when deal health degrades. * **Personalized outreach at scale**: The Outreach Email Drafting skill generates AI-personalised emails with A/B variants tailored to prospect role, industry, and intent signals, in under 2 minutes per prospect. * **Centralized knowledge access**: The Persona AI queries Confluence, Google Drive, and prior submissions to draft RFP responses, surface battlecards, and handle objections with consistent positioning. ## How the Persona AI helps you get things done faster The Persona AI accelerates core tasks by eliminating manual data gathering and document assembly. The following examples show the time savings across common activities. | Task | Without Persona AI | With Persona AI | | ------------------------ | ---------------------------------------------------- | ---------------------------------------------------------- | | Account brief generation | 30-60 min of manual research across multiple systems | Under 5 min, auto-generated from connected data | | Proposal creation | 2-4 hours of compiling data and formatting | Under 30 min, generated from templates and live data | | Prospect research | Variable, manual lookups across tools | Under 2 min per prospect | | RFP response drafting | Days of searching prior submissions and writing | 60%+ faster with AI-assisted drafting from knowledge bases | | Deal risk assessment | Periodic manual review, often too late | Continuous, with proactive alerts on degrading deal health | | Follow-up email drafting | 15-30 min per personalised email | Under 2 min with context from CRM and call data | The Persona AI removes the repetitive data gathering that consumes most of the working day, so your team spends time on decisions and relationships instead of copying between systems. # Mortgage Insurance Source: https://docs.ideaboxai.com/use-cases/enact-mortgage-copilot Learn how IdeaBoxAI's Mortgage Insurance Persona AI transforms data discovery, underwriting decisions, and customer support for a private mortgage insurer, connecting Actian DI, Snowflake, and Google Drive into a single persona-aware assistant. Every ML engineer at a mortgage insurer knows the dataset hunt. You need to build a credit scoring model. There are 47 tables in Snowflake that look relevant. Three of them have "customer" in the name. One is certified, one is a staging copy, and one is a raw extract that nobody owns. The metadata catalog tells you the tables exist, but not which one to trust, whether it is current, or what is missing for your use case. You spend two days figuring out which data you can actually use before writing a single line of model code. Every underwriter knows the tab-switching. A mortgage application is on your desk. You need the applicant's risk factor from Snowflake, the LTV coverage rules from the policy guide, and the pipeline status to see how many other applications are waiting. Three systems, three logins, three different interfaces, just to make one approval decision. Every customer support agent knows the hold time. A borrower calls asking about their coverage. You need to look up their policy, check their claim status, and confirm the cancellation rules, all while the customer waits. If the answer is not in the first document you open, you put them on hold and start searching. The Mortgage Insurance Persona AI solves all three. It connects Actian DI (Zeenea) for metadata intelligence, Snowflake for live structured data, and Google Drive for policy documentation, unified into a single persona-aware assistant accessible from the browser and Microsoft Teams. ## The knows your role before you ask A Persona in IdeaBoxAI is the AI's understanding of who you are, what you need, and what your data looks like. When you log in as an ML Engineer, you get a data intelligence layer tuned to dataset discovery, certification status, and lineage. When you log in as an Underwriter, you get a decisioning layer tuned to application pipelines, risk factors, and policy rules. When you log in as a Customer Support Agent, you get a resolution layer tuned to coverage questions, claim status, and escalation summaries. The Mortgage Insurance Persona AI ships with three purpose-built personas. Each connects to the data sources relevant to their role. The persona you are assigned to determines which data you see, which skills are available, and which scenario cards appear on your home screen. The following table summarises the three personas. | **Persona** | **Focus** | **Key Needs** | | ---------------------- | ------------------------------------ | ---------------------------------------------------------------------------------------------------------- | | Data and ML Engineer | Datasets, models, data quality | Certified dataset discovery, table disambiguation, model gap analysis, lineage tracking, NL-to-SQL queries | | Underwriter | Applications, risk, policy rules | Application pipeline visibility, risk filtering, eligibility decisions, policy lookups, pipeline aging | | Customer Support Agent | Policies, claims, borrower questions | Coverage answers, claim status, cancellation rules, escalation summaries | Each persona also comes with pre-configured scenario cards on the home screen. Click any card to pre-fill the chat input with a common workflow prompt. You can also type your request directly in plain language; the Persona AI auto-detects the right skill. To browse all available skills, type `/` in the chat bar to open the skill picker. ## ML Engineer scenario: dataset discovery for model building **Anika Patel ยท Senior ML Engineer** Credit risk team ยท Building a new credit scoring model ยท Needs certified training data Anika has been asked to build a new credit scoring model. Before she can start, she needs to find which datasets are available, which ones are certified for production use, and whether any required features are missing. Before IdeaBoxAI, this meant opening the Atacama catalog, searching for relevant tables, then cross-referencing Snowflake to check freshness and row counts, a process that usually took 1โ€“2 days. **Anika types:** *"Which datasets are certified and trusted for building a credit scoring model?"* **The Persona AI responds:** > **Certified dataset discovery, credit scoring** > 47 tables in Snowflake match credit-related metadata ยท 8 are certified for production use ยท 3 are directly relevant to credit scoring > > **Certified datasets for credit scoring** > > | **Table** | **Schema** | **Certification** | **Owner** | **Last updated** | **Row count** | > | -------------------------- | ----------------- | ----------------- | ------------------- | ---------------------------- | ------------- | > | CREDIT\_SCORE\_FEATURES | PROD.RISK | โœ… Certified | Risk Analytics Team | Apr 22, 2026 (daily refresh) | 2.4M | > | LOAN\_APPLICATION\_HISTORY | PROD.UNDERWRITING | โœ… Certified | Underwriting Ops | Apr 22, 2026 (daily refresh) | 8.1M | > | BORROWER\_PROFILE | PROD.CUSTOMER | โœ… Certified | Data Engineering | Apr 21, 2026 (daily refresh) | 3.2M | > > **Why these three:** > > * CREDIT\_SCORE\_FEATURES contains the core feature set used in the current production model, FICO scores, DTI ratios, employment history, and 14 derived risk indicators. > * LOAN\_APPLICATION\_HISTORY provides outcome labels (approved, denied, defaulted) across 8.1M historical applications. Essential for supervised training. > * BORROWER\_PROFILE links applicant demographics and LTV ratios to each application record. > > **Potentially relevant but not certified** > > | **Table** | **Schema** | **Issue** | **Recommendation** | > | -------------------------- | -------------- | -------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | > | RAW\_CREDIT\_BUREAU | STAGING.RISK | Not certified, raw extract from credit bureau feed | Do not use for model training. This is the upstream source for CREDIT\_SCORE\_FEATURES. Use the certified downstream table instead. | > | CUSTOMER\_PAYMENT\_HISTORY | PROD.SERVICING | Certified for servicing, not for credit scoring | Contains 36 months of payment behaviour. Relevant for churn prediction but not labelled for credit scoring. Request cross-certification from the Risk Analytics team if you need it. | > > **Data lineage note:** CREDIT\_SCORE\_FEATURES is derived from RAW\_CREDIT\_BUREAU through 3 transformation stages. The most recent pipeline run completed at 2:14 am today. No quality alerts in the last 30 days. > > **Suggested follow-ups:** Show me the full column list for CREDIT\_SCORE\_FEATURES ยท What features are missing for a churn prediction model ยท Trace the lineage for BORROWER\_PROFILE ยท Run a sample query on LOAN\_APPLICATION\_HISTORY Anika follows up with a question that used to require opening Slack and asking the data engineering team. **Anika types:** *"Is BANK\_CUSTOMER the authoritative table or should I use RAW\_CUSTOMER or STAGING\_CUSTOMER?"* **The Persona AI responds:** > **Table disambiguation, CUSTOMER tables** > 3 tables match "customer" ยท 1 is authoritative > > | **Table** | **Schema** | **Status** | **Refresh** | **Row count** | **Verdict** | > | ----------------- | ---------------- | ------------------------------- | ------------- | ------------- | --------------------------------------------------------------------------------------------------------------------------------------------------- | > | BANK\_CUSTOMER | PROD.CUSTOMER | โœ… Certified ยท Authoritative | Daily, 3:00am | 3.2M | **Use this one.** This is the certified production table. Owned by Data Engineering. | > | STAGING\_CUSTOMER | STAGING.CUSTOMER | โš ๏ธ Not certified ยท Staging copy | Hourly | 3.2M | Do not use. This is the staging table that feeds BANK\_CUSTOMER. Data may include incomplete records from the current day's ingest. | > | RAW\_CUSTOMER | RAW\.CUSTOMER | โŒ Not certified ยท Raw extract | On ingest | 4.1M | Do not use. Raw extract from the source system. Contains 900K records that failed validation and were filtered out in the staging-to-prod pipeline. | > > **Key difference:** RAW\_CUSTOMER has 4.1M rows vs BANK\_CUSTOMER's 3.2M. The 900K gap is intentional, those records failed data quality checks (missing SSN, invalid state codes, or duplicate applicant IDs). If your model trains on RAW\_CUSTOMER, it will include invalid records that the production pipeline already rejected. > > **Lineage:** Source system โ†’ RAW\_CUSTOMER โ†’ STAGING\_CUSTOMER โ†’ BANK\_CUSTOMER. The BANK\_CUSTOMER table is the end of the validated pipeline. Training a model on uncertified tables risks embedding data quality issues into production predictions. The Persona AI surfaces certification status and lineage so engineers can make informed decisions, not assumptions. The results: * Certified datasets identified in seconds, with ownership, freshness, and row counts from the live metadata catalog. * Table disambiguation resolved without asking the data engineering team on Slack. The 900K record gap between raw and production tables explained with specific validation reasons. * 1โ€“2 days of dataset hunting reduced to a 3-minute conversation. ## Underwriter scenario: application review and eligibility **James Wu ยท Senior Underwriter** Reviews 15โ€“20 mortgage applications per day ยท Pipeline review with manager at 11 am James reviews mortgage applications for approval. Each decision requires checking the applicant's risk profile in Snowflake, confirming eligibility against policy rules in the coverage guide, and tracking how long the application has been in review. Before IdeaBoxAI, this meant three browser tabs and a spreadsheet to track aging. **James types:** *"Show me all pending mortgage applications waiting for my approval."* **The Persona AI responds:** > **Application pipeline, James Wu** > 18 applications pending ยท 3 flagged for attention > > **๐Ÿ”ด Priority, aging or high risk** > > | **Application** | **Applicant** | **Loan amount** | **LTV** | **Credit score** | **Risk factor** | **Days in review** | **Flag** | > | --------------- | --------------- | --------------- | ------- | ---------------- | --------------- | ------------------ | ---------------------------------------------------------------------------- | > | APP-28471 | Martinez, Elena | \$340,000 | 92% | 618 | 14.2 | **8 days** | Aging, exceeds 5-day SLA. Risk factor above threshold. | > | APP-28455 | Chen, Robert | \$520,000 | 88% | 642 | 12.8 | **6 days** | Aging, exceeds 5-day SLA. | > | APP-28502 | Johnson, Tanya | \$275,000 | 95% | 580 | 16.1 | 2 days | Highest risk factor in queue. Requires senior review per policy section 4.2. | > > **๐ŸŸก Standard review, 12 applications** > All within SLA. Risk factors between 4.0 and 10.0. Average days in review: 2.4. > > | **Risk range** | **Count** | **Average loan amount** | **Average LTV** | > | ------------------- | --------- | ----------------------- | --------------- | > | 4.0โ€“6.0 (low risk) | 5 | \$310,000 | 78% | > | 6.1โ€“8.0 (moderate) | 4 | \$385,000 | 84% | > | 8.1โ€“10.0 (elevated) | 3 | \$420,000 | 89% | > > **๐ŸŸข Ready for fast-track, 3 applications** > All three have risk factors below 4.0, LTV under 80%, and credit scores above 740. These qualify for automated approval under policy section 2.1. > > **Suggested follow-ups:** Show me the full details for APP-28471 ยท Check eligibility for APP-28502 ยท How many applications this week have a risk factor above 12 ยท Show me my pipeline aging trend over the last 30 days James clicks into the highest-risk application. **James types:** *"I have an applicant with a 580 credit score and 10% down, are they eligible?"* **The Persona AI responds:** > **Eligibility decision, 580 credit score, 10% down payment** > > | **Criteria** | **Applicant** | **Policy requirement** | **Status** | > | -------------- | ------------- | ---------------------------------- | --------------------- | > | Credit score | 580 | Minimum 620 for standard approval | โŒ Below threshold | > | Down payment | 10% (LTV 90%) | Minimum 5% required | โœ… Meets requirement | > | LTV ratio | 90% | Maximum 97% for insured loans | โœ… Within limit | > | Debt-to-income | Not provided | Maximum 45% per policy section 3.4 | โš ๏ธ Needs verification | > > **Eligibility verdict: Not eligible for standard approval.** > > The 580 credit score is 40 points below the minimum threshold of 620 defined in policy section 3.1 (Minimum Credit Requirements). This is a hard floor, it cannot be waived at the underwriter level. > > **Alternative paths:** > > * If the applicant has compensating factors (12+ months of reserves, no derogatory marks in 24 months), the application may qualify for exception review under policy section 3.1.3. Exception reviews require VP-level sign-off. > * If the credit score is within 90 days of a rescore event (dispute resolution, paid collection), recommend the applicant reapply after the rescore. The current application should be placed on hold, not denied. > > **Policy source:** Coverage Guide, Section 3.1, Minimum Credit Requirements. Last updated: January 2026. > > **Suggested follow-ups:** Show me the full policy on exception reviews ยท Check if this applicant has compensating factors ยท What is the maximum LTV ratio for a 620 credit score ยท Pull the DTI requirements for this loan amount The Persona AI pulls eligibility rules from the policy guide in Google Drive and cross-references applicant data from Snowflake. Every answer cites the specific policy section so the underwriter can verify the source. The results: * 18 applications surfaced with risk scores, aging status, and SLA flags, no spreadsheet tracking needed. * Eligibility decision returned in seconds with the specific policy section, threshold, and alternative paths. * 3 applications auto-identified for fast-track approval, freeing underwriter time for the complex cases. * Pipeline aging visible at a glance, with applications exceeding the 5-day SLA flagged automatically. ## Underwriter scenario: risk filtering before pipeline review **James Wu ยท Senior Underwriter** Pipeline review with manager in 30 minutes ยท Needs to know which applications carry the highest risk James needs to walk into his pipeline review with a clear picture of risk concentration. His manager will ask how many high-risk applications are in the queue and what the team is doing about them. **James types:** *"How many applications this week have a RISK\_FACTOR above 12?"* **The Persona AI responds:** > **Risk filtering, RISK\_FACTOR above 12** > Week of April 21, 2026 ยท 74 applications received ยท 6 have a risk factor above 12 > > | **Application** | **Applicant** | **Risk factor** | **Credit score** | **LTV** | **Loan amount** | **Days in review** | **Assigned to** | > | --------------- | --------------- | --------------- | ---------------- | ------- | --------------- | ------------------ | --------------- | > | APP-28502 | Johnson, Tanya | 16.1 | 580 | 95% | \$275,000 | 2 | James Wu | > | APP-28471 | Martinez, Elena | 14.2 | 618 | 92% | \$340,000 | 8 | James Wu | > | APP-28519 | Okafor, David | 13.8 | 625 | 91% | \$410,000 | 1 | Sarah Lin | > | APP-28487 | Patel, Neha | 13.1 | 630 | 88% | \$295,000 | 4 | James Wu | > | APP-28510 | Reeves, Mark | 12.6 | 640 | 90% | \$380,000 | 2 | Sarah Lin | > | APP-28455 | Chen, Robert | 12.8 | 642 | 88% | \$520,000 | 6 | James Wu | > > **Risk concentration analysis:** > > * 6 of 74 applications this week (8.1%) are above the 12.0 threshold. This is slightly above the trailing 4-week average of 6.8%. > * 4 of the 6 are assigned to James Wu. Consider redistributing APP-28487 or APP-28455 to balance the high-risk review load. > * The common pattern: all 6 have LTV above 88% and credit scores below 645. The risk is driven by thin equity combined with borderline credit, not a single outlier factor. > > **SLA status:** 2 of the 6 (APP-28471 and APP-28455) have exceeded the 5-day review SLA. Both are assigned to James. > > **Suggested follow-ups:** Show me the risk factor trend over the last 4 weeks ยท Check eligibility for APP-28502 ยท Redistribute my high-risk applications ยท Compare this week's risk profile to the same week last quarter James walks into his review with the exact numbers: 6 high-risk applications, 8.1% concentration (above the 4-week average), two past SLA, and a clear pattern linking thin equity to borderline credit. ## Support scenario: coverage question and claim lookup **Diana Reyes ยท Customer Support Agent** Handles 40โ€“50 borrower calls per day ยท Current call: borrower asking about coverage Diana is on a call with a borrower who wants to know what their mortgage insurance covers and whether they can cancel it early. Before IdeaBoxAI, she would put the borrower on hold, open the coverage guide PDF, search for the relevant section, and hope the search term matched. **Diana types:** *"What is the coverage for a homebuyer who put 15% down?"* **The Persona AI responds:** > **Coverage lookup, 15% down payment** > > A homebuyer who put 15% down has a loan-to-value (LTV) ratio of 85%. The following coverage applies. > > | **Coverage attribute** | **Detail** | > | ------------------------------- | ----------------------------------------------------- | > | LTV at origination | 85% | > | MI coverage percentage | 12% of the loan amount | > | Monthly premium (typical range) | 0.30%โ€“0.55% of the loan amount annually, paid monthly | > | Coverage type | Borrower-paid mortgage insurance (BPMI) | > > **What the coverage means in practice:** > If the borrower defaults and the lender files a claim, mortgage insurance covers 12% of the original loan amount. On a $300,000 loan, that is $36,000 in coverage to the lender. > > **Cancellation eligibility:** > > * The borrower can request cancellation when LTV reaches 80% (through principal paydown or home appreciation). This requires a current appraisal. > * Automatic cancellation occurs when LTV reaches 78% based on the original amortization schedule, per the Homeowners Protection Act. > > **Policy source:** Coverage Guide, Section 2.3, Standard Coverage Percentages by LTV Band. Last updated: January 2026. > > **Suggested follow-ups:** At what LTV ratio does MI automatically cancel ยท Can a borrower cancel early with a new appraisal ยท What is the coverage for a 10% down payment ยท Look up a specific policy number The borrower asks a follow-up. Diana does not need to put them on hold. **Diana types:** *"Can a borrower cancel their mortgage insurance early and if so, how?"* **The Persona AI responds:** > **Cancellation rules, early cancellation** > > A borrower can cancel mortgage insurance early through two paths. > > **Path 1: Borrower-requested cancellation (LTV reaches 80%)** > > | **Requirement** | **Detail** | > | --------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------- | > | LTV threshold | 80% or below | > | How LTV is calculated | Based on current principal balance vs original property value, or a new appraised value if the borrower requests a reappraisal | > | Payment history | Must be current on payments. No 30-day late payments in the last 12 months. No 60-day late payments in the last 24 months. | > | Process | Borrower submits a written request to the servicer. Servicer orders appraisal if needed. Cancellation takes effect on the first day of the month following approval. | > > **Path 2: Automatic cancellation (LTV reaches 78%)** > > | **Requirement** | **Detail** | > | ------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------- | > | LTV threshold | 78% based on the original amortisation schedule | > | No borrower action needed | The servicer is required to cancel MI automatically on the date the principal balance is scheduled to reach 78% of the original property value. | > | Legal basis | Homeowners Protection Act of 1998 (HPA) | > > **Key distinction:** Borrower-requested cancellation at 80% can use a new appraisal to account for home value appreciation. Automatic cancellation at 78% uses only the original amortization schedule; home appreciation is not factored in. > > **What to tell the borrower:** "You can request cancellation now if your balance is at or below 80% of your home's current value. If you believe your home has appreciated, you can request a reappraisal through your servicer. Otherwise, your MI will automatically cancel when your scheduled balance reaches 78% of the original value, no action needed on your end." > > **Policy source:** Coverage Guide, Section 6.1, Cancellation and Termination. Homeowners Protection Act of 1998. > > **Suggested follow-ups:** Look up a specific borrower's current LTV ยท When will MI auto-cancel for a specific loan number ยท What is the reappraisal process ยท Explain the difference between BPMI and LPMI Diana answers the borrower's question in real time, no hold, no document search, no transfer to a senior agent. **Diana types:** *"What is the current status of claim number 48291?"* **The Persona AI responds:** > **Claim status, #48291** > > | **Field** | **Detail** | > | ------------------- | ------------------------------------------------------- | > | Claim number | 48291 | > | Policy number | PMI-2024-07-38291 | > | Borrower | Williams, Karen | > | Loan amount | \$285,000 | > | Claim filed | March 14, 2026 | > | Claim type | Default, 90+ days delinquent | > | Status | **Under review**: assigned to Claims Analyst Team B | > | Documents received | 3 of 4 required (missing: updated property valuation) | > | Expected resolution | Within 15 business days of complete document submission | > > **What is holding this claim:** The updated property valuation has not been received. This document was requested from the servicer on March 18. It has been 26 business days with no submission. > > **Recommended action:** Contact the servicer to follow up on the outstanding property valuation. Once received, the claim moves to final review and should resolve within 15 business days. > > **Suggested follow-ups:** Draft a follow-up email to the servicer about the missing valuation ยท Show me all claims for this borrower ยท Summarise this claim for escalation to a senior agent The Persona AI pulls claim data from Snowflake and policy details from Google Drive in a single query. Support agents never need to switch systems, the answer arrives with the source cited. The results: * Coverage question answered in seconds with the specific policy section, premium range, and cancellation rules, no PDF searching. * Borrower-facing language provided alongside the technical answer so the agent can respond naturally. * Claim status pulled with the specific blocker identified (missing document, 26 days outstanding) and the next action ready. * Average handle time reduced by eliminating hold time, system switching, and manual document lookup. ## Skills that power all three personas Each persona in IdeaBoxAI is powered by a set of Skills, AI capabilities that define how the Persona AI handles specific types of requests. The Mortgage Insurance Persona AI ships with 15 skills across the three personas. You can trigger a skill in three ways: * Click a **scenario card** on the home screen to pre-fill the prompt and run the skill immediately. * Type your request in plain language. The Persona AI auto-detects the appropriate skill from your message. * Type `/` in the chat bar to open the skill picker and select a skill by name. You do not need to memorize skill names or commands. Type what you need in plain language and the Persona AI matches your intent to the right skill automatically. ### Data and ML Engineer skills The following skills handle dataset discovery, data quality validation, and analytical queries against the warehouse. | **Skill** | **What it does** | | --------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Dataset Discovery | Searches the Actian DI (Zeenea) catalog for certified, trusted datasets that match a use case. Returns certification status, owner, freshness, and row counts. | | Table Disambiguation | Compares similarly named tables and identifies the authoritative source. Surfaces lineage, certification, and the differences between raw, staging, and production copies. | | Model Gap Analysis | Maps available datasets against the requirements for a specific model type (credit scoring, churn, risk). Identifies what exists, what is missing, and who owns each gap. | | NL-to-SQL | Translates natural language questions into SQL queries against Snowflake. Returns results directly, bypassing the catalog for live structured queries. | | Lineage and Ownership | Traces a table's full data lineage from source system to production. Shows transformation stages, refresh schedules, and the team responsible for each step. | ### Underwriter skills The following skills handle application reviews, risk assessment, and policy-based eligibility decisions. | **Skill** | **What it does** | | -------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Application Pipeline | Surfaces all pending mortgage applications assigned to the underwriter. Flags aging, risk level, and SLA status per application. | | Risk Filtering | Queries Snowflake for applications matching risk criteria, risk factor thresholds, credit score ranges, or LTV bands. Returns counts, distributions, and trends. | | Policy Q\&A | Answers policy-related questions by querying the coverage guide in Google Drive. Returns the answer with the specific section cited. | | Eligibility Decision | Evaluates an applicant's profile against policy eligibility criteria. Returns a pass/fail verdict per criterion with the specific policy threshold and alternative paths. | | Pipeline Aging | Identifies applications that have exceeded review SLAs. Ranks by days outstanding and flags patterns in aging concentration. | ### Customer Support Agent skills The following skills handle borrower-facing questions, claim lookups, and escalation preparation. | **Skill** | **What it does** | | ------------------ | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Coverage Q\&A | Answers coverage questions based on loan characteristics (LTV, down payment, loan type). Returns coverage percentages, premium ranges, and cancellation eligibility. | | Policy Q\&A | Answers policy and regulatory questions from the coverage guide. Returns the answer with borrower-facing language and the source section cited. | | Claim Status | Looks up a claim by number or borrower name. Returns the current status, assigned team, outstanding documents, and expected resolution timeline. | | Cancellation Rules | Explains the paths to mortgage insurance cancellation, borrower-requested and automatic, with requirements, process steps, and legal basis. | | Escalation Summary | Assembles a structured summary of an account dispute or complex case, pulling data from Snowflake and policy context from Google Drive, ready to hand off to a senior agent. | ## Connected to your data infrastructure The Mortgage Insurance Persona AI sits on top of three data sources, unified into a single persona-aware assistant. It does not replace your existing infrastructure, it makes it accessible through natural language. When an ML engineer asks "which datasets are certified for credit scoring?", the Persona AI queries the Actian DI (Zeenea) catalog for certification status and lineage, then cross-references Snowflake for freshness and row counts. The answer arrives in seconds, with the specific tables, their owners, and the reasons to use or avoid each one. Each persona only sees the data relevant to their role. An ML engineer queries the metadata catalog and Snowflake. An underwriter queries Snowflake application data and Google Drive policy documents. A support agent queries Snowflake claim records and Google Drive coverage guides. The platform connects through the following layers: * Actian DI (Zeenea) connected via API, loads metadata context, data lineage, and certification status from the enterprise catalog. * Snowflake connected as a structured knowledge base, NL-to-SQL queries run directly against the warehouse for live application, claim, and risk data. * Google Drive connected via RAG pipeline, retrieval-augmented Q\&A over policy documents, coverage guides, and underwriting manuals. * IdeaBoxAI browser app for all personas, dedicated web interface with full conversational and dashboard capabilities. * Microsoft Teams bot integration, Teams-native queries for users who work inside Teams throughout the day. ## Getting started Setting up the Mortgage Insurance Persona AI for your team takes less than a day. Link Actian DI (Zeenea) via API, Snowflake as a structured knowledge base, and Google Drive as a document source from the Connections settings in the Admin Console. Each integration is configured and tested to confirm data visibility. Set up Data and ML Engineer, Underwriter, and Customer Support Agent personas in the Admin Console. Assign the relevant skills and knowledge base to each. The knowledge base scopes each persona to their relevant data sources automatically. Click the knowledge base generation button. The platform maps your Snowflake schema, indexes your Google Drive policy documents, and connects to the Actian DI catalog. This takes under 10 minutes. Add your team members and assign them to their persona. They log in and see their role-specific Persona AI immediately. Scenario cards guide them through the most common workflows from day one. Start on the IdeaBoxAI dev environment to validate connections and test persona responses against your real data. Once the team is confident, promote to production. The Mortgage Insurance Persona AI is available now as part of IdeaBoxAI's Persona AI suite. Contact the IdeaBoxAI team to set up a live demo connected to your data infrastructure. ## Next steps Explore these guides to learn more about the platform capabilities behind the Mortgage Insurance Persona AI. Create and configure role-specific personas in the Admin Console. Assign skills to personas and customize how the Persona AI handles requests. Connect structured data sources and generate knowledge bases for grounded responses. Understand the measurable outcomes the Conversational AI delivers. # Grounded responses with Knowledge Bases Source: https://docs.ideaboxai.com/use-cases/grounded-responses See how connecting live business data to the Persona AI produces accurate, context-aware answers instead of generic responses. ## Overview This use case demonstrates how a financial services firm used IdeaBoxAI's Knowledge Bases to transform their AI-powered customer support from generic, hallucination-prone responses into precise, data-grounded answers โ€” reducing customer complaint escalations by 62% and improving first-contact resolution rates from 54% to 89%. ## Background **Horizon Financial Services**, a mid-market investment advisory firm managing portfolios for over 3,200 clients, deployed an AI chatbot to handle routine customer inquiries about account balances, transaction history, investment performance, and fee structures. Initially, they used a standard large language model trained on general financial knowledge. While the AI could answer broad questions like *"What is a diversified portfolio?"* or *"How does dollar-cost averaging work?"*, it consistently failed when customers asked specific questions about their own accounts, such as: * *"What was my total return on my growth portfolio last quarter?"* * *"Why was I charged a \$45 fee in March?"* * *"Which of my holdings have dividends scheduled this month?"* The AI would either provide generic responses (*"Fee structures vary by account type โ€” please review your agreement"*), make up plausible-sounding but incorrect information (hallucination), or deflect to human support. This created frustration for customers who expected instant, accurate answers and overwhelmed the human support team with avoidable escalations. ## The challenge Horizon Financial Services faced three critical issues with their initial AI implementation: 1. **Hallucinated responses** โ€” Without access to real customer data, the AI would occasionally fabricate account details, transaction amounts, or fee explanations that sounded credible but were factually incorrect. This eroded customer trust and created compliance risk. 2. **Generic deflection** โ€” For most account-specific questions, the AI would respond with unhelpful generic statements like *"Please contact your advisor for details"* or *"You can view this information in your account portal"* โ€” which defeated the purpose of having an AI assistant in the first place. 3. **High escalation rate** โ€” 46% of AI conversations escalated to human agents within the first three interactions because the AI couldn't answer basic, data-driven questions. This created a bottleneck in the support queue and negated the cost savings the AI was supposed to provide. ## The solution Horizon Financial Services implemented IdeaBoxAI's Knowledge Base system to connect their customer account database (PostgreSQL) and transaction ledger (MySQL) directly to their AI Persona, enabling the AI to query live data and ground every response in real, verified information. ### Implementation approach The IT team created two Knowledge Bases in IdeaBoxAI: * **Customer Accounts KB** โ€” Connected to their PostgreSQL database containing account types, balances, portfolio allocations, asset holdings, and advisor assignments. * **Transaction History KB** โ€” Connected to their MySQL ledger with all deposits, withdrawals, trades, dividends, and fee transactions dating back 5 years. Each Knowledge Base was configured with a semantic data model that mapped database columns (e.g., `acct_balance`, `txn_type`, `fee_code`) to natural language concepts the AI could understand. The team created a custom AI Persona in IdeaBoxAI called **Horizon Assistant** with clear instructions: * Always query the Knowledge Bases before responding to account-specific questions. * Cite the specific data source (account number, transaction ID, date) in responses. * If data is not available in the Knowledge Bases, explicitly state *"I don't have access to that information"* rather than guessing. * For compliance-sensitive topics (e.g., tax advice, legal guidance), deflect to human advisors with clear disclaimers. The Persona was attached to both Knowledge Bases, enabling it to execute SQL queries across customer accounts and transaction history in real time. To ensure data privacy and compliance, the team configured row-level security in the Knowledge Bases: * Customers could only query data associated with their authenticated account ID. * Advisors could access data for all accounts they managed. * The AI was restricted from executing any `UPDATE`, `DELETE`, or `INSERT` operations โ€” read-only access only. The team also implemented query timeout limits (5 seconds) and result set caps (100 rows) to prevent performance issues. Before launching to customers, the support team tested the AI with 200+ real customer questions collected from historical support tickets. They refined the Knowledge Base semantic layer, added missing data mappings, and adjusted the Persona's instructions to handle edge cases (e.g., closed accounts, pending transactions, currency conversions). ### Key configurations Horizon Financial Services leveraged several IdeaBoxAI features to optimize accuracy and compliance: * **Semantic data modeling** โ€” Instead of forcing the AI to write raw SQL, the Knowledge Base abstracted database schema into business terms. For example, the AI could understand *"What are my dividend-paying stocks?"* and automatically map it to `SELECT symbol FROM holdings WHERE dividend_yield > 0 AND account_id = [user]`. * **Citation and transparency** โ€” Every AI response included the data source and timestamp. For example: *"Your growth portfolio returned +8.3% last quarter (Q1 2026, data from transaction ledger as of May 7, 2026)."* This transparency increased customer confidence. * **Fallback handling** โ€” When the AI encountered ambiguous queries or incomplete data, it asked clarifying questions (*"I see three portfolios under your account โ€” which one are you asking about: Growth, Income, or Balanced?"*) rather than guessing. * **Human handoff triggers** โ€” The Persona was configured to automatically escalate to a human agent if it detected keywords related to complaints, legal issues, or emotional distress โ€” ensuring sensitive situations were handled appropriately. ## Results Within 120 days of deploying grounded Knowledge Base responses, Horizon Financial Services achieved significant improvements in support quality and efficiency: | Metric | Before | After | Improvement | | ------------------------------------ | -------------- | --------------- | ----------- | | **First-contact resolution rate** | 54% | 89% | +35 pts | | **Customer complaint escalations** | 112/month | 43/month | -62% | | **Average resolution time** | 18 minutes | 3.5 minutes | -81% | | **Hallucinated/incorrect responses** | 11% of queries | \<1% of queries | -91% | | **Customer satisfaction (CSAT)** | 3.2/5 | 4.6/5 | +44% | | **Support ticket volume** | 2,400/month | 980/month | -59% | ### Qualitative outcomes Beyond the quantitative metrics, the team and customers reported several qualitative benefits: * **Customer trust and confidence** โ€” Customers appreciated receiving precise, data-backed answers with clear citations. Comments like *"This is exactly what I needed โ€” no runaround"* and *"Better than talking to a human advisor"* became common in feedback surveys. * **Reduced compliance risk** โ€” By eliminating hallucinated financial advice, Horizon significantly reduced their exposure to regulatory penalties and customer disputes stemming from incorrect AI-generated information. * **Support team morale** โ€” Human agents spent less time answering repetitive, data-lookup questions and more time on high-value advisory work โ€” leading to higher job satisfaction and lower turnover. * **Scalability without headcount** โ€” The firm onboarded 400+ new clients during the pilot period without increasing support headcount. The grounded AI absorbed the additional query volume seamlessly. ## Key takeaways This use case highlights three core capabilities of grounded AI responses powered by Knowledge Bases: 1. **Factual accuracy over plausibility** โ€” By querying live databases instead of relying on pre-trained model knowledge, the AI eliminated hallucinations and delivered verifiable, source-cited answers every time. 2. **Data privacy and access control** โ€” Row-level security and read-only query permissions ensured that customers could only access their own data, and the AI couldn't accidentally modify or expose sensitive information. 3. **Semantic abstraction for non-technical queries** โ€” The Knowledge Base's semantic layer allowed customers to ask questions in plain English (*"What did I earn last month?"*) without needing to understand database schemas, SQL syntax, or technical jargon. ## Technical considerations For teams evaluating grounded AI implementations, Horizon's experience surfaced several important technical lessons: * **Semantic modeling is critical** โ€” Invest time upfront mapping database columns to natural language business concepts. A well-modeled Knowledge Base reduced incorrect queries by 73% compared to raw SQL access. * **Query performance matters** โ€” Slow queries (>5 seconds) caused customers to abandon conversations. Horizon optimized their database indexes and implemented query caching to maintain sub-2-second response times for 95% of queries. * **Iterate with real data** โ€” Testing with synthetic or example questions is insufficient. Real customer queries exposed edge cases, ambiguous phrasing, and missing data mappings that synthetic tests missed. * **Explicit "I don't know" is better than guessing** โ€” Configuring the AI to admit when data is unavailable increased trust more than attempting to provide partial or inferred answers. ## Related resources Learn how to connect a live database and build a semantic data layer for AI queries. Understand how structured Knowledge Bases work with SQL databases like PostgreSQL and MySQL. Create custom AI assistants and attach Knowledge Bases for grounded responses. Learn how to connect your Knowledge Base to a Persona AI. # Use cases Source: https://docs.ideaboxai.com/use-cases/index See how teams across industries use IdeaBoxAI to transform their workflows with AI-powered personas, skills, and connected data. Explore real deployments across construction, technology sales, financial services, and more. Each case study covers persona design, skill configuration, sample Persona AI interactions, and data integration details. AP invoice triage, job cost monitoring, RFI tracking, and daily log generation. 3 personas, 24 skills, connected to Sage 300 CRE via ODBC. Account research, deal health scoring, technical discovery debriefs, and outbound prospecting. 3 personas, 32 skills, connected to Salesforce, Confluence, and Google Drive. Dataset discovery, underwriting pipeline management, risk filtering, and customer support. 3 personas, 15 skills, connected to Actian DI, Snowflake, and Google Drive. Problem-to-capability mapping, efficiency benchmarks, and success metrics for measuring Persona AI impact across sales workflows. Agentic BI applied to a real supply chain and sales scenario to drive faster, data-informed decisions. How connecting live business data to the Persona AI produces accurate, context-aware answers. How teams use IdeaBoxAI Automations to generate and distribute reports without manual effort. # Construction & Real Estate Source: https://docs.ideaboxai.com/use-cases/sage-cre-copilot Learn how IdeaBoxAI's Construction & Real Estate Persona AI transforms AP, project management, and field engineering workflows in construction and real estate, powered by 24 purpose-built skills across three personas. Every AP Coordinator knows the Friday morning feeling. Forty-seven open invoices. A cash position that won't cover all of them. Three vendors calling about payment status. A lien deadline on Monday that nobody flagged until now. Every Project Manager knows the equivalent. A client wants an owner report by 3 pm. You have nine active jobs. Three of them are probably over budget. You just don't know which three yet, or by how much, or why. Every Project Engineer knows the daily grind. Five cost codes are burning faster than schedule percentage. Two RFIs are overdue with no response. A subcontractor's insurance expired last week, and nobody noticed. The Sage Persona AI solves all three, not with another report or another dashboard, but with a Persona AI that connects directly to your Sage 300 CRE data via ODBC. It understands your role and answers your actual questions in plain English. ## Getting started Setting up the Sage Persona AI for your team takes less than a day. Link your Sage environment from the Connections settings in the platform. The ODBC connector is configured with a read-only service account on your Sage 300 server network. No data leaves your network; only summarised query results are sent to the AI layer. Set up Project Manager, Project Engineer, and AP Coordinator personas in the Admin Console. Assign the relevant skills and knowledge base to each. The knowledge base scopes each persona to their relevant Sage tables automatically. Click the knowledge base generation button. The platform uses AI to map your Sage table structure, understand the relationships between modules, and generate the cubes needed for dashboards. This takes under 10 minutes. Add your team members and assign them to their persona. They log in and see their role-specific Persona AI and dashboard immediately. Starter prompts guide them through the most common tasks from day one. Start on the IdeaBoxAI dev environment to validate the connection and test the persona responses against your real data. Once the team is satisfied, promote to production. The Sage Persona AI is available now as part of IdeaBoxAI's Persona AI suite. Contact the IdeaBoxAI team to set up a live demo connected to your Sage 300 CRE environment. ## The Persona AI knows your role before you ask A Persona in IdeaBoxAI is the AI's understanding of who you are, what you need, and what your data looks like. The persona you are assigned to determines which data you see, which skills are available, and which starter prompts appear on your home screen. The Sage Persona AI ships with three purpose-built personas. Each connects to the Sage 300 CRE tables relevant to their role via ODBC. | **Persona** | **Focus** | **Key Needs** | | ---------------- | --------------------------------------------------- | --------------------------------------------------------------------- | | Project Manager | Portfolio P\&L, owner reporting, change orders | Budget vs actual, profit forecast, schedule risk, cash flow | | Project Engineer | Job cost by cost code, RFIs, subcontract compliance | Cost code burn rates, overdue RFIs, change order drafting, daily logs | | AP Coordinator | Invoices, payments, vendor compliance | PO matching, payment prioritisation, W-9 and insurance tracking | ## AP scenario: payment run prioritization **Priya Sharma - AP Coordinator** 12 active projects, 47 open invoices, cash position review due by 9 am. Priya types: *"Which open commitment invoices should I pay first this week, and are any of them at lien risk?"* **The Persona AI responds** with a payment priority summary covering all 47 invoices ranked by lien risk, early-pay discount expiry, due date, and PO match status: > **Payment priority summary, week of April 17:** > 47 open invoices across 12 projects. \$1.24M total outstanding. Cash available: \$940,000. > > **Action required today:** > > * **ABC Electrical, INV-2301** (\$84,200): Lien deadline Mon Apr 21. No waiver on file. Preliminary notice filed Feb 3. > * **XYZ Mechanical, INV-2289** (\$12,400): 2% early-pay discount expires today at 5 pm. PO matched, no holds. > * **ABC Plumbing, INV-2294** (\$31,000): Lien deadline Thu Apr 24. Waiver requested Mar 28, no response in 20 days. > > **Recommended payment run:** 6 invoices totaling \$183,200. All have approved POs, no compliance holds. > > **Cash position after full run:** \$629,200 remaining. > **38 remaining invoices:** \$929,200 total. None due before Apr 28. Priya follows up: *"Show me all invoices from ABC Electrical across all jobs."* The Persona AI returns a vendor summary with 4 invoices across 2 projects, identifies \$115,600 in lien-exposed invoices with no waivers on file, and recommends requesting conditional lien waivers before releasing payment. **The results:** * 47 invoices ranked by risk in under 2 minutes. * \$115,600 in lien-exposed invoices flagged with vendor history. * Friday morning prep reduced from 90 minutes to a 5-minute review. ## AP scenario: daily morning brief **Marcus Okafor - AP Coordinator** 8 active construction projects, 3 states, 31 open invoices Marcus types: *"Give me my AP morning brief, what needs my attention today?"* **The Persona AI responds** with a structured brief covering action items, payments due, and a week-ahead outlook: > **AP morning brief, Thursday, April 17, 2026** > 8 active projects. 31 open invoices. \$1.24M total AP liability > > **Action required today:** > > * XYZ HVAC, INV-4410 (\$18,400, Midtown Tower): No matching PO found. First mismatch for this vendor across 6 prior invoices. > * ABC Framing: Insurance cert expires Mon Apr 21. Active on 2 projects. Send renewal request today. > > **Due in 48 hours, ready to process:** > 3 invoices totaling \$62,800. All PO matched, no holds, waivers on file. > > **Heads up, next 7 days:** > XYZ MEP insurance expires in 11 days. 4 invoices totaling \$87,400 due between Apr 22-25. Marcus reviews the brief in under 3 minutes, flags XYZ HVAC for investigation, and approves the payment run before his 9 am stand-up. ## PM scenario: portfolio health scan **Sarah Chen - Project Manager** 9 active jobs, portfolio review with VP at 10 am Sarah types: *"Scan my portfolio and flag any jobs where cost-to-complete is trending over original budget."* **The Persona AI responds** with a portfolio scan covering all 9 jobs: > **Portfolio health scan, April 17, 2026** > 9 active jobs. \$82.4M total committed value. 3 flagged > > **Harborview Mixed-Use (high risk):** Original budget \$16.4M, committed \$18.2M (+11%). Contingency effectively exhausted at \$42,000. Driven by \$1.1M structural steel overrun plus 3 unsigned change orders (\$184,000). Overrun accelerating from 4% at 50% complete to 11% at 60%. > > **Eastside Warehouse (watch):** 6% over budget at 45% complete. \$480,000 in mechanical extras from late architectural revision. Contingency at 3.6%, runs out at \~65% complete at current pace. > > **Ridgecrest Office (marginal):** 2% over at 72% complete. Minor extras across three trades. Likely closes within contingency. > > **6 other jobs:** All tracking within 1% of original budget. Sarah follows up: *"Show me the change order status for Harborview."* The Persona AI returns the full change order summary showing \$184,000 in unsigned COs, including one that is 18 days overdue (2x the normal turnaround with this owner), and recommends specific escalation actions. **The results:** * 9 jobs health-scored in seconds with trend analysis and contingency burn rate. * \$184,000 in revenue risk surfaced with specific next steps. * Monday morning prep reduced to a 4-minute conversation. ## PM scenario: owner report generation **David Osei - Project Manager** Client report due at 3 pm, normally takes 90 minutes David types: *"Generate an owner progress report for Lakeside Residences. Include budget vs actual, committed costs, schedule status, and top risks."* **The Persona AI generates** a structured report with an executive summary, budget and cost summary, schedule status by phase, pending change orders requiring owner action, and top risks with decision deadlines: > **Lakeside Residences, Owner Progress Report** > > **Executive summary:** Project is 64% complete, tracking within 1.5% of original budget. Framing is 4 days behind baseline due to weather delay. Two items require owner decisions this week. > > **Key figures:** Original contract \$18.4M, revised \$18.55M, projected final \$18.61M (+0.3% variance). Contingency remaining: \$248,000 (1.3%). > > **Top risks:** > > * Window package order deadline Apr 24. 14-week lead time means any delay adds to critical path. > * CO-009 facade changes (\$148,000) due Apr 22. Facade sub cannot mobilize until executed. **The results:** * Full owner progress report generated in under 10 minutes instead of 90. * Data pulled automatically from four Sage modules. ## PE scenario: cost code burn rate alert **Jake Torres - Project Engineer** 2 assigned jobs, 6:45 am site arrival Jake types: *"Which cost codes are burning faster than the schedule percentage on my jobs?"* **The Persona AI responds** with a burn rate analysis across both jobs: > **Cost code burn rate alert, April 17, 2026** > 2 jobs. 47 active cost codes. 5 flagged > > **Harborview Mixed-Use (3 codes at risk):** > > * 03-200 Structural concrete: 72% burned at 60% schedule (+12%). Projected \$148,000 over. > * 09-100 Drywall: 62% burned at 48% schedule (+14%). Labour rate running 8% above estimate. > * 15-400 HVAC ductwork: 67% burned at 55% schedule (+12%). Ductwork rerouting from architectural revision. > > Combined projected overrun: \$305,600 across 3 cost codes. > > **Eastside Warehouse (2 codes to watch):** Both within 5% variance. Flag for weekly monitoring. Jake follows up: *"Draft a change order for the extra excavation work on Eastside."* The Persona AI generates a CO document with actual rates from JCCD cost data, schedule impact assessment, and contract references, ready for PM review. **The results:** * 5 at-risk cost codes identified with root cause analysis. * Change order drafted with actual rates in under 60 seconds. * Morning site prep reduced from 30 minutes to 3 minutes. ## PE scenario: RFI tracking and daily log **Maria Santos - Project Engineer** 3 assigned jobs, end of day on site Maria types: *"Which RFIs are overdue for a response on my jobs?"* **The Persona AI responds** with an overdue RFI tracker: > **Overdue RFI tracker, April 17, 2026** > 3 jobs. 14 open RFIs. 4 overdue > > * **RFI-042** (Harborview, beam depth discrepancy): 10 days overdue. Stop-work risk if not resolved by Apr 21. > * **RFI-038** (Eastside, electrical panel clearance): 5 days overdue. Blocks electrical rough-in. > * **RFI-051** (Northgate, roof drain location): 3 days overdue. > * **RFI-053** (Harborview, fire rating): 2 days overdue. Maria fires off follow-up emails, then types: *"Generate my daily construction log for today."* The Persona AI generates a structured daily log pulling crew data from PMPRJ and PMSUB records, work order completions, equipment on site, weather, and issues, ready for review and submission. ## PM scenario: cash flow forecasting **Elena Vasquez - Project Manager** MEP subcontractor, 5 active jobs, payment timing is critical Elena types: *"Show me my cash flow forecast for the next 30 days across all active jobs."* **The Persona AI responds** with a 30-day forecast: > **30-day cash flow forecast, April 17 to May 17, 2026** > 5 active jobs. Net 30-day position: +\$213,000. Timing risk in week 3 > > Expected inflows: \$847,000 across 5 jobs (confidence scored per client payment history). > Expected outflows: \$634,000 (subs, materials, equipment, insurance). > > **Timing risk, April 24 to May 7:** > A 14-day window where cumulative cash flow goes negative (-\$66,000 peak). > The XYZ Medical payment (\$312,000) is expected Apr 22 but this client > averages 34 days on their last 8 payments. If they pay on their actual > pattern, the gap widens to -\$140,000 for 4 days. > > **Recommendation:** Follow up with XYZ Medical AP today to confirm > payment date. Consider accelerating the Lakeside draw or deferring the > HVAC equipment payment as fallback. **The results:** * \$140,000 cash flow gap identified 10 days before it would have hit. * Payment confidence scored per inflow based on actual client history. * Proactive follow-up with a concrete fallback plan. ## Skills that power all three personas Each persona is powered by a set of skills that define how the Persona AI handles specific types of requests. The Sage Persona AI ships with 24 skills across the three personas, 8 per role. You can trigger a skill in two ways: * Click a **starter prompt** on the home screen to pre-fill the chat input. * Type your request in plain language. The Persona AI auto-detects the appropriate skill. You do not need to memorize skill names or commands. Type what you need in plain language, and the Persona AI matches your intent to the right skill automatically. ### Project Manager skills | **Skill** | **What it does** | | ---------------------- | ------------------------------------------------------------------------------- | | Portfolio Risk Scanner | Scans JCJOB and JCCD records. Ranks jobs by budget variance. | | Profit Forecast | Projects current margin vs at-completion margin per job. | | Owner Report Generator | Assembles progress, cost tables, change orders, and open items into a report. | | Change Order Tracker | Monitors pending change orders by age. Alerts on overdue approvals. | | Schedule Risk Analyser | Calculates float remaining and identifies critical-path delays. | | Cash Flow Forecast | Projects billings vs committed costs over the next 60 days. | | Morning Briefing | Returns top 5 priorities: overruns, stale COs, overdue RFIs, upcoming billings. | | Document Drafter | Generates owner emails, CO follow-ups, and schedule recovery summaries. | ### Project Engineer skills | **Skill** | **What it does** | | ------------------------ | --------------------------------------------------------------------------- | | Job Cost Monitor | Monitors cost codes burning ahead of schedule percentage. | | Budget Overrun Alert | Scans for jobs where actuals exceed budget on assigned jobs. | | RFI Tracker | Tracks open and overdue RFIs. Alerts on missed response deadlines. | | RFI Drafter | Generates structured RFI documents with contract context. | | Change Order Drafter | Creates CO documents from scope and actual JCCD cost data. | | Subcontractor Compliance | Scans for expired insurance and missing lien waivers. | | Daily Log Generator | Generates structured daily logs from crew data and work orders. | | Morning Briefing | Returns top 5 issues: overruns, overdue RFIs, compliance gaps, pending COs. | ### AP Coordinator skills | **Skill** | **What it does** | | ---------------------- | ---------------------------------------------------------------------------- | | Invoice Approval Queue | Pulls pending invoices sorted by aging. | | Over-PO Detector | Flags invoices that exceed their purchase order amount. | | Compliance Checker | Scans for missing W-9s, expired insurance, absent lien waivers. | | Payment Run Builder | Prioritises invoices by due date, discount window, and liability. | | AP Liability Reporter | Aggregates unpaid invoices by job with due-date breakdown. | | Miscoding Detector | Flags job and property codes that do not match active records. | | Vendor Contact Drafter | Drafts W-9 requests, insurance renewal emails, payment confirmations. | | Morning Briefing | Returns top 5 AP priorities: aging invoices, over-PO flags, compliance gaps. | ## Connected to your Sage 300 CRE data The Sage Persona AI connects to your live Sage data via a read-only ODBC service account. No ETL, no nightly sync, no data warehouse required. Each persona only sees the data relevant to their role. An AP Coordinator queries APINV, APVD, JCPO, and PMSUB. A Project Manager queries JCJOB, JCCD, CTJOB, CTCO, and PMPRJ. A Project Engineer queries JCCD, PMRFI, PMSUB, and PMPRJ. Key details: * 10 Sage 300 modules accessed: JCJOB, JCCD, JCPO, CTJOB, CTCO, PMPRJ, PMRFI, PMSUB, APINV, and APVD. * Role-scoped knowledge base per persona. * Dashboards for visual reporting: PM Overview, PE Overview, and AP Overview. * Gmail integration for one-click email drafting. Draft-first, user reviews before sending. * Data queries return in under 4 seconds. Document generation in under 8 seconds. ## Next steps Create and configure role-specific personas in the Admin Console. Assign skills to personas and customize how the Persona AI handles requests. Connect structured data sources and generate knowledge bases for grounded responses. Understand the measurable outcomes the Persona AI delivers. # Supply chain and sales Source: https://docs.ideaboxai.com/use-cases/supply-chain-sales Learn how Agentic BI was applied to a real supply chain and sales scenario to drive faster, data-informed decisions. ## Overview This use case demonstrates how a mid-sized distribution company used IdeaBoxAI's Agentic BI to unify fragmented supply chain and sales data, enabling real-time visibility into inventory levels, order fulfillment rates, and regional sales performance โ€” ultimately reducing stockouts by 34% and improving forecast accuracy. ## Background **Metro Distribution Partners**, a regional wholesale distributor serving over 450 retail locations across the Midwest, faced a common but costly challenge: their supply chain and sales teams operated in silos, each relying on separate spreadsheets and static monthly reports. Inventory planners worked from Excel exports pulled weekly from their ERP system, while the sales team tracked orders and customer trends in a separate CRM database. This fragmentation led to frequent miscommunication. Sales would promise delivery timelines based on outdated inventory data, while supply chain teams would overstock slow-moving SKUs and understock high-demand products because they lacked visibility into real-time sales velocity and regional trends. ## The challenge Metro Distribution Partners struggled with three core issues: 1. **Data fragmentation** โ€” Inventory data lived in their ERP system, sales orders in the CRM, and logistics metrics in a third-party warehouse management platform. No single dashboard provided a unified view. 2. **Manual reporting bottleneck** โ€” The operations manager spent 8โ€“10 hours each week manually consolidating CSV exports, building pivot tables, and emailing static PDF reports to regional managers. By the time the reports were distributed, the data was already outdated. 3. **Reactive decision-making** โ€” Without real-time alerts, the team only discovered stockouts or fulfillment delays during weekly review meetings โ€” too late to prevent customer impact or lost revenue. ## The solution Metro Distribution Partners implemented IdeaBoxAI's Agentic BI to connect their ERP database (PostgreSQL), CRM system (MySQL), and warehouse logistics data (CSV exports) into a single, AI-generated dashboard ecosystem. ### Implementation approach The team created two Knowledge Bases in IdeaBoxAI: one connected to their ERP database (PostgreSQL) containing SKU-level inventory, restocking schedules, and supplier lead times; and another connected to their CRM database (MySQL) with sales orders, customer accounts, and regional sales rep assignments. Weekly logistics reports (CSV) were uploaded directly to Agentic BI for trend analysis. Using Agentic BI's auto-generation mode, the operations manager created three core dashboards: * **Inventory Health Dashboard** โ€” Real-time stock levels, reorder alerts, and days-of-supply metrics by SKU and warehouse location. * **Sales Performance Dashboard** โ€” Revenue by region, top-performing SKUs, order fulfillment rates, and sales velocity trends. * **Supply Chain Overview** โ€” Order cycle time, supplier lead time variances, and stockout frequency by product category. Each dashboard included drill-through interactions. For example, sales managers could click on a regional revenue card and instantly drill through to see individual sales reps, then further drill down to view SKU-level order details. Supply chain planners could roll up SKU-level inventory into category-level summaries or drill down to warehouse-specific stock levels. The team configured email alerts for critical thresholds: * Notify inventory planners when any SKU drops below 7 days of supply. * Alert the operations manager when order fulfillment rate falls below 92%. * Notify regional sales managers when a top-10 SKU goes out of stock in their territory. ### Key configurations Metro Distribution Partners took advantage of several Agentic BI features to tailor the dashboards to their workflow: * **AI Assistant for chart refinement** โ€” Instead of manually reconfiguring charts, the operations manager used plain English commands like *"Show me inventory by supplier and highlight SKUs with lead times over 14 days"* or *"Add a stacked bar chart of weekly order volume by region."* * **Canvas grouping** โ€” Dashboards were organized into collapsible sections: **KPIs** (top-line metrics), **Inventory Analysis** (stock levels and reorder status), **Sales Trends** (revenue and velocity), and **Logistics Performance** (fulfillment and lead times). * **Dashboard filters** โ€” Regional managers could filter the entire dashboard by their assigned territory, product category, or date range โ€” enabling each stakeholder to view only the data relevant to their role. * **Embedded dashboards** โ€” The Supply Chain Overview dashboard was embedded into the company's internal operations portal using an iframe, giving warehouse staff instant access without needing to log into IdeaBoxAI separately. ## Results Within 90 days of deploying Agentic BI, Metro Distribution Partners achieved measurable operational improvements: | Metric | Before | After | Improvement | | -------------------------------------------- | ------------- | ----------- | ----------- | | **Stockout incidents per month** | 47 | 31 | -34% | | **Order fulfillment rate** | 89% | 96% | +7 pts | | **Forecast accuracy** | 72% | 89% | +17 pts | | **Time spent on manual reporting** | 8โ€“10 hrs/week | 30 min/week | -92% | | **Average response time to stockout alerts** | 3.5 days | 4 hours | -96% | ### Qualitative outcomes Beyond the quantitative metrics, the team reported several qualitative benefits: * **Cross-functional alignment** โ€” Sales and supply chain teams now worked from the same real-time dashboards during weekly planning meetings, eliminating discrepancies and finger-pointing. * **Faster decision cycles** โ€” Regional managers could answer critical questions like *"Which SKUs are trending up this month?"* or *"Do we have enough inventory to support the Q4 promotion?"* in seconds, not days. * **Proactive inventory management** โ€” Automated alerts enabled the inventory planning team to reorder stock before stockouts occurred, rather than reacting to customer complaints after the fact. ## Key takeaways This use case highlights three core capabilities of Agentic BI that are especially valuable in supply chain and sales contexts: 1. **Unified data view** โ€” By connecting multiple data sources (ERP, CRM, logistics CSVs) through Knowledge Bases, Metro Distribution Partners eliminated data silos and gained a single source of truth. 2. **AI-accelerated dashboard creation** โ€” Auto-generation mode reduced the time to build a functional dashboard from weeks (for traditional BI tools) to under 10 minutes, allowing the team to iterate quickly and test different views. 3. **Proactive alerting** โ€” Real-time threshold alerts transformed the team from reactive firefighters into proactive planners, enabling them to prevent issues before they impacted customers or revenue. ## Related resources Learn how to connect your data sources and generate your first dashboard. Configure automated email notifications for critical metric thresholds. Enable stakeholders to filter dashboards by region, category, or date range. Step-by-step tutorial for creating a sales dashboard from your data.