Define custom research insights for sales agents

By default, the company research insights generated by sales agents such as Sales Qualification Agent and Sales Opportunity Agent cover key areas such as company overview, financials, and recent news. However, you can enhance the agent's research capabilities by adding custom research insights tailored to your sales team's specific needs.

For example, you can configure the agent to provide insights about recent conversations with the contact, key stakeholders and decision makers, past spends based on internal billing data, or any other relevant information that can help your sellers better understand and engage with your customers. Learn more in the Example scenarios section.

Add custom research insights and knowledge sources

The configuration steps to add custom research insights are similar for both the sales agents.

1. Go to App Settings > Dynamics 365 AI Hub > Create and manage agents and then select the sales agent you want to configure (for example, Sales Qualification Agent).

2. Under Knowledge, select Research, and then scroll down to the Create custom research section.

3. Select Create custom research and enter the following details:

   • Display name: Enter a name for the custom research insight. This name is used as a title for the research insights generated by the agent and shown on the research page of the lead or opportunity.

   • What insight you want the seller to see: Describe the type of insight you want the agent to generate. This instruction acts as a prompt for the agent to generate relevant research insights. For example, "Summarize the lead company's past purchase history from our internal billing system."

   • Preferred knowledge source: Select only one knowledge source or MCP tool per custom research insight. Review the Best practices for adding custom research insights section. Depending on the agent you're configuring, the knowledge source or MCP tool you select here is associated with the following Copilot Studio agents:

     • Sales Qualification Agent: D365 Sales Agent - Custom Research
     • Sales Close Agent: Sales Close Agent - Custom Research

     Perform ONE of the following steps:

     • Select an existing knowledge source. The agent lists the knowledge sources and MCP tools that you already added to the corresponding Copilot Studio agent.

     • Add a new knowledge source or MCP tool.

       • Select Add in Copilot Studio from the drop-down.
         This action opens the corresponding Copilot Studio agent in a new browser tab.

       • Select Add knowledge to add a Public URL, SharePoint document, Azure Data Lake, and Dataverse table as the knowledge source. To add an MCP tool or connector to non-Microsoft sources, select the Tools tab and then select Add a tool. Learn more in Add tools to custom agents.

       • Return to the sales agent settings page and select the newly added source from the drop-down.

     
     If you don't select any knowledge source, the agent relies on its own knowledge, public data, and other sources you configured.

4. Select Add to create the custom research insight.

5. Select the Test icon ( ) in the Custom research section.

6. Follow the instructions in the Test insight pane and verify whether the agent is generating relevant research insights using the knowledge sources you added. If the response isn't satisfactory, refine the prompt in the What insight do you want the seller to see field and retest.

7. After the test is successful, open the corresponding Copilot Studio agent and select Publish.

8. Return to the sales agent settings page, save the changes, and start the agent.
   The agent will generate the custom account research insights for new records and records that meet the reprocessing criteria.
   Sellers can view these insights under the Deeper insights section of the lead or opportunity research page. The following screenshot shows the custom research insights added to Lead research page by Sales Qualification Agent:

   

Best practices for adding custom research insights

When you add custom research insights to sales agents, follow these best practices to ensure high-quality and relevant insights for your sellers:

• Start with a single, high-value internal data source to validate quality before scaling.
• Use precise and consistent naming conventions for each insight, such as QBR Summary, Competitor Comparison, or Partner Risk Assessment.
• Avoid overloading the agent with multiple or loosely relevant data sources, as this approach can reduce output accuracy.
• Regularly evaluate the effectiveness of each custom insight based on seller feedback and usage telemetry.
• When you use a non-public knowledge source or an MCP tool, ensure that sellers have the necessary access permissions to these sources. This access enables them to open the citation links and cross-verify the insights if needed.

Example scenarios

This section provides example scenarios that illustrate the types of custom research insights you can add to the sales agents by using different knowledge sources and MCP tools.

Example: Custom insights about recent conversations and engagements

Let's say that your sellers want to get the context of any recent conversations, calls, and meetings that took place between the customer/prospect and people from your company. And imagine your sales organization uses Gong for its powerful revenue insights. You can use the Gong's Microsoft Graph connector to synthesize those insights as part of the sales agent's research using the following steps:

1. Link Gong to your Microsoft 365 graph connector by using the Microsoft 365 admin center. Learn more in Gong Copilot connector. While setting up the connector, keep the following points in mind:

   • Ensure that you have sufficient access in the Gong app to view the access key.
   • Ensure that your Dynamics 365 Sales environment is in the same Microsoft 365 tenant where you set up the Gong graph connector.
   • Select Custom setup and set the User permissions to Everyone so that your sellers get the complete context of ongoing conversations, otherwise they only have access to conversations they directly had in the past with the lead’s company.

   After Gong successfully starts syncing with Microsoft Graph, you can see the Gong connector as an available knowledge source in Copilot Studio.

2. Follow the steps in Add custom research insights and knowledge sources section to add Gong as a knowledge source, test it, and publish the agent.
   When the agent researches a lead or opportunity, it provides insights about recent calls. The following screenshot shows an example of research insights generated by Sales Qualification Agent using Gong as a knowledge source:

Example: Custom insights about stakeholder and decision-making process

You're setting custom research insights for your company Contoso Inc that produces and sells GPUs to other companies. To effectively engage with your customers and prospects, your sellers need three additional insights:

The key stakeholders and decision makers – Who are the right stakeholders and decision makers for the company? The company’s 2025 AI roadmap – Where is the company in their AI adoption roadmap? Are they running any AI workloads? The company’s tooling posture - Summarize the recent service case history for the company.

Let's see how your sellers can get additional insights about key stakeholders and decision makers in your lead's company. In this example, we'll use a connector to a non-Microsoft data source that provides firmographic and technographic data about companies called 'Enlyft.'

1. Create a new custom research insight in the agent settings page.

2. Add the following details:

   • Title: Key stakeholders and decision makers

   • Insights you want the seller to see: Who are the key decision makers and stakeholders for the company?

   • Preferred knowledge source: Select Add in Copilot Studio from the drop-down and select the Enlyft connector tool that you already configured in Copilot Studio.

3. Test the custom research insight and publish the agent. When the agent researches a lead, it provides insights about key stakeholders and decision makers in the lead's company.

Similarly, you can create additional custom research insights for the company's AI roadmap and tooling posture by adding relevant prompts and knowledge sources in the agent settings page as shown in the following screenshot:

Other example scenarios

For custom research, use available knowledge sources and connectors or Model Context Protocol (MCP) Servers—including those from leading technology vendors in the CRM domain. Set up your custom knowledge sources and tools in Copilot Studio, and link them to the sales agent to generate tailored research insights for your sellers.

Some examples of what you can include:

• Get more context about the company from briefing notes that your sales team uploads onto a SharePoint site.

• Summarize orders, service cases, and any other tables from Microsoft Dataverse.

• Use a host of available knowledge sources in Microsoft Copilot Studio.

• Get a wide range of company insights from leading data providers such as Enlyft and Draup by using their connectors (available through the Tools menu in Copilot Studio). For example:

  • Get company tech stack (Draup)

  • Get IT spend by category (Draup)

  • Get key stakeholders or business objectives (Enlyft)

  • Get account plan overview (Enlyft)

• Get insights from any applications or systems of record you use by adding your connector, prompt, or MCP Server into Copilot Studio. Learn more in Create and add a new tool at agent level.

Related information

• Lead research page
• Opportunity research page