Share via

Add unstructured data as a knowledge source

This article provides information to add unstructured data, such as OneDrive, SharePoint, and Power Platform connectors as knowledge sources for your agent. For more information about unstructured data, go to Unstructured data as a knowledge source.

Add OneDrive files and folders

You can add files and folders from OneDrive for use as a knowledge source for your agent. The differences between adding files and folders using OneDrive versus uploading files:

  • Using the upload files method only allows you to select individual files. With OneDrive, you can add both files and folders.
  • The upload files method requires that you reload the file if changes were made to the file. In the upload files method, files are static. So, if the file is updated, those updates aren't reflected in the uploaded version unless manually updated. With OneDrive, files and folders that are added are synchronized, so changes are updated automatically.
  • The upload files method allows anyone with permission to the agent to access the file. With OneDrive, the user's credentials are verified to ensure they have permission to access the file before providing responses.

When you use OneDrive as a knowledge source, your files are stored in Dataverse, where they're processed and indexed for use. Your files and folders remain in Dataverse, so they do consume data storage.

OneDrive as a knowledge source currently supports the following file types:

  • Word: .doc, .docx
  • PowerPoint: .ppt, .pptx
  • PDF: .pdf
  • Excel: .xls, .xslx

The maximum file size is 32 MB.

Documents that were protected using sensitivity labels, or were password protected, can't be indexed. If added, these types of documents show as ready for use but don't provide responses.

  1. Open the agent.

  2. Select Add knowledge from either the Overview or Knowledge pages, or the Properties of a generative answers node.

  3. From the Add knowledge dialog, select OneDrive.

  4. Select Browse items to choose specific files or folders, or enter the URL of a OneDrive file.

    1. When you select Browse items, you can select up to five individual files, folders, or combinations of the two. For more information about limits, go to OneDrive limits.

    2. Select Confirm selection once you selected all of your items.

    3. Specify a URL for items that you have shared access to, but aren't stored in your OneDrive.

  5. After you added your items from OneDrive, you can update the name and description in Copilot Studio. The description should be detailed, especially if generative AI is enabled, because it helps generative orchestration. Changes to the name in Copilot Studio don't affect the files or folders in OneDrive.

  6. Select Add to finish adding the OneDrive items to your knowledge source.

  7. On the Knowledge page, you can view the status of the OneDrive items you added. The amount of time it takes to index your items depends on the number of files, the size of the files, and type of files. Also, if this is the first time that you're configuring this data source, extra time is required to create the Dataverse schema.

    Note

    Individual files appear as individual items in the list. A folder is represented as one item, and includes all its subfolders and files.

  8. Once the status of your items is set to Ready, you can ask your agent questions in the Test panel.

  9. The first time you ask questions in the Test panel, you have to connect to verify your permissions to access the content.

Add SharePoint files and folders

You can add files and folders from SharePoint for use as a knowledge source for your agent. The differences between adding files and folders using SharePoint versus uploading files:

  • Using the upload files method only allows you to select individual files. With SharePoint, you can add both files and folders.
  • The upload files method requires that you reload the file if changes were made to the file. In the upload files method, files are static. So, if the file is updated, those updates aren't reflected in the uploaded version unless manually updated. With SharePoint, files and folders that are added are synchronized, so changes are updated automatically.
  • The upload files method allows anyone with permission to the agent to access the file. With SharePoint, the user's credentials are verified to ensure they have permission to access the file before providing responses.
  • Currently, document libraries are not supported.

When you use SharePoint as a knowledge source, your files are stored in Dataverse, where they're processed and indexed for use. Your files and folders remain in Dataverse, so they do consume data storage.

SharePoint as a knowledge source currently supports the following file types:

  • Word: .doc, .docx
  • PowerPoint: .ppt, .pptx
  • PDF: .pdf
  • Excel: .xls, .xslx

The maximum file size is 32 MB.

Documents that were protected using sensitivity labels, or were password protected, can't be indexed. If added, these types of documents show as ready for use but don't provide responses.

  1. Open the agent.

  2. Select Add knowledge from either the Overview or Knowledge pages, or the Properties of a generative answers node.

  3. From the Add knowledge dialog, select SharePoint.

  4. Select Browse items to choose specific files or folders, or enter the URL of a SharePoint file.

    1. When you select Browse items, you can select up to five individual files, folders, or combinations of the two. For more information about limits, go to SharePoint limits.

    2. Select Confirm selection once you selected all of your items.

    3. Specify a URL for items that you have shared access to, but aren't stored in your SharePoint.

  5. After you add your items from SharePoint, you can update the name and description in Copilot Studio. The description should be detailed, especially if generative AI is enabled, because it helps generative orchestration. Changes to the name in Copilot Studio don't affect the files or folders in SharePoint.

  6. Select Add to finish adding the SharePoint items to your knowledge source.

  7. On the Knowledge page, you can view the status of the SharePoint items you added. The amount of time it takes to index your items depends on the number of files, the size of the files, and type of files. Also, if this is the first time that you're configuring this data source, extra time is required to create the Dataverse schema.

    Note

    Individual files appear as individual items in the list. A folder is represented as one item, and includes all its subfolders and files.

  8. Once the status of your items is set to Ready, you can ask your agent questions in the Test panel.

  9. The first time you ask questions in the Test panel, you have to connect to verify your permissions to access the content.

Add knowledge base data

Knowledge bases are a key way organizations share information, both for internal users and customers. They often span sales, service, support, policy, and more. However, knowledge base content is often unstructured. This unstructured, nontabular storage is often less expensive than relational, table-based storage, while also allowing for more dynamic configuration than tables.

Copilot Studio currently supports the following unstructured sources:

  • Confluence (cloud only)
  • Salesforce
  • ServiceNow
  • ZenDesk

Unstructured knowledge base content is added at the "collection" level. This content might be called "Pages", "Knowledge Bases," or some other term in the individual source, but in this context, think of them as "folders." You can't select individual files.

Only published articles are used. Draft or archived content isn't used.

User credentials are used to validate that the user has permission to use the content. Before providing a summary or answer, the agent verifies permissions.

  1. Open the agent.

  2. Select Add knowledge from either the Overview or Knowledge pages, or the Properties of a generative answers node.

  3. From the Add knowledge dialog, select the unstructured knowledge base source, Confluence, Salesforce, ServiceNow, or ZenDesk.

    Note

    If your admin already configured a Microsoft 365 connection to the knowledge base source, the source is listed in the Featured section. Otherwise, check the Advanced section.

  4. Select the Your connections drop-down and select Create new connection. At the prompt, sign in to your knowledge base source.

    Note

    The connection that is created during this setup allows the maker to access the data to create the knowledge source for the agent. These credentials aren't used by the agent's users after the agent is published. Those users have to sign in with their own credentials.

  5. Select Allow to provide Copilot Studio access to your knowledge base source, then select Next.

  6. Select your knowledge base source from the list of tables and sources available in your environment.

    Note

    The information icon in the source provides a message to indicate that the source is unstructured.

  7. Select Preview to review your data before it's retrieved. A sample set of the knowledge base source is displayed, and you can select individual rows to view the content you're going to use.

  8. With your knowledge base source selected, select Next.

  9. Update the name and description of the knowledge base source. The description should be detailed, especially if generative AI is enabled, because it helps generative orchestration. Changes to the name in Copilot Studio don't affect the files or folders in the source files.

  10. Select Add to complete the process.

  11. On the Knowledge page, you can view the status of the knowledge base source you added. The amount of time it takes to index your items depends on the number of files, the size of the files, and type of files. Also, if this is the first time that you're configuring this data source, extra time is required to create the Dataverse schema.

  12. Once the status of your items is set to Ready, you can ask your agent questions in the Test panel.

  13. The first time you ask questions in the Test panel, you have to connect to verify your permissions to access the content.