Add Dataverse tables as agent knowledge sources in Copilot Studio.
When integrating Dataverse tables as a knowledge source, you will rely on the data contained in those tables to support your agent. This process includes adding synonyms and defining terms for the tables and columns in your data.
Note :
- To use this knowledge base, Copilot Studio agents must use Dataverse search. If you cannot add Dataverse tables to the agent, ask your administrator to enable Dataverse search in your environment.
- Using Dataverse tables as a knowledge source requires you to set agent authentication to Authentication with Microsoft. The other two options, No authentication and Authentication manually , are not supported.
To add Dataverse tables as a knowledge source, follow these steps:
1. Open the agent.
2. Select Add knowledge from the Overview or Knowledge page .
3. Select Dataverse .
4. Find one or more of your Dataverse tables to add. You can add up to 15 Dataverse tables per knowledge source. To narrow down your options, use the search field.
Note : Table suggestions are based on your agent name.
5. Review the knowledge name and description. Provide a detailed description, especially if Generative AI is enabled, as descriptions aid in generative coordination.
6. Optionally, to help improve the quality of the answer, add synonyms and glossary terms:
- Select Edit and add synonyms to the table columns you have selected. Select the Back button to accept the changes.
- Select Edit and add glossary terms to define specialized terminology and abbreviations. Select the Back button to accept the changes.
7. Select Add to agent to complete adding the knowledge source.
Synonyms and Glossary of Terms
Synonyms, glossary terms, and definitions for synonym and glossary entries help to govern AI. They provide baseline data for improving generated responses. This additional information helps AI understand and interpret the data in your tables. It also increases the AI's ability to recognize user requests and return responses based on that information.
For cases where your Dataverse table contains a column of numerical values, you need to provide a synonym so the AI understands the content of that column. For example, your agent is providing travel assistance services, and your Dataverse table contains a column named "cr_123_abc" that uses flight numbers to correspond to cities.
Since AI doesn't know how to categorize this information, you have to specify how to interpret it. Therefore, the creator adds a description to this column, such as the following example: "cr_123_abc represents the departure city of each flight indicated by the flight code."
Examples of term definitions
Use term definitions to rephrase terms in your Dataverse, helping agents better understand user queries and respond more effectively.
The following table illustrates how adding definitions to terms in the glossary provides useful context for your agent.
| Script | Terminology | Sample description |
|---|---|---|
| Abbreviations | VP | "VP" refers to the Vice President value in the "JobTitle" column of the "Contact" table. |
| Custom ownership | activity owner | The "PartyId" column in the "ActivityParty" table identifies the "activity owner". |
| Custom field | potential revenue | "Opportunity revenue" refers to the "Custom Revenue" column in the "Opportunity" table. |
| Complex rules or filters | overdue task | "Overdue task" refers to the "task" table when the "state code" column has a value of open and the "scheduled end date" column has a value earlier than today. |
Note :
- The descriptions in the table are for illustrative purposes only. To verify which description yields the best results, test your agent.
- It may take up to 15 minutes for the terms and definitions in the glossary to be updated.
Enable search support for multi-line text and file data types in the Dataverse table (preview)
[This section is pre-release material and is subject to change.]
When adding Dataverse as a knowledge source, you can apply unstructured inference to obtain higher-quality responses from Multiline Text ( MemoType) and File ( FileType) columns.
Important note :
- This article contains preview documentation for Microsoft Copilot Studio and is subject to change.
- Preview features are not intended for use in a production environment and may have limited functionality. These features are provided before the official release so you can access them early and provide feedback.
- If you're building an agent ready for a production environment, check out the Microsoft Copilot Studio Overview .
Before performing a search on the Dataverse, ensure you meet the following requirements:
- Dataverse's search capability has been enabled.
- You have creator or administrator access (necessary to modify views in Power Apps).
Important note : Creating a search index will incur additional Dataverse storage costs.
Configure the Dataverse table in Power Apps.
For this preview feature, in Quick Find View , explicitly include the table, and the Multiline Text and File columns as Searchable .
- Log in to Power Apps and select your preferred environment.
- Select Dataverse , then select Tables .
- Select the Dataverse table you added to the Copilot Studio agent.
- Enable Searchable for each column you want to search.
- In the Data experiences pane , select Views .
- From the list of views, select the Quick Find View type .
- Select a searchable column from the list and add it to the view.
- Select Edit find tables columns to add columns that can be searched using the Find by options .
- Select Save and Publish to publish the changes to the view.
Known limitations
- If you add the Dataverse knowledge source before configuring Multiline Text and File columns in Power Apps, the system may take up to two days to repopulate the request. To speed things up, consider adding the Dataverse knowledge source again after configuring the search field.
- Tables, images, and text in languages other than those of the organization are not supported in attached files.
- When you add Dataverse virtual tables as a knowledge source, only virtual tables associated with the Finance and Operations data provider are available for selection. Any other data provider is not supported.
- Enhance agent knowledge in Copilot Studio
- Add public websites as knowledge sources for agents in Copilot Studio.
- Check the Copilot Studio agent.
- Overview of Copilot Studio
- Should you choose Microsoft 365 Copilot or Copilot Studio to build your agent?
- Evaluating agent performance in Microsoft Copilot Studio
- Add SharePoint as a knowledge source for agents in Copilot Studio.
- Overview of AI-based agent creation