Add Copilot Control to a canvas app (preview)

[This article is prerelease documentation and is subject to change.]

The Copilot control is a next-generation AI assistant that makers can add to their canvas apps for end-users. This is an AI-powered experience for app users to get insights about the data in their apps through conversation in natural language. Makers can add this control to any canvas app and choose what data it can answer questions about.

Important

  • To use this capability your environment must be in the US region.
  • Preview features aren’t meant for production use and may have restricted functionality. These features are available before an official release so that customers can get early access and provide feedback.
  • For more information, go to our preview terms.
  • This capability is powered by Azure OpenAI Service.
  • This capability may be subject to usage limits or capacity throttling.
  • Copilot control isn't supported and won't work for environments that have customer-managed key (CMK) or have lockbox.

Step 1 - Enable Copilot for your environment

In order for end users to use the Copilot chat experience in a canvas app, an admin must first turn on the feature, Allow users to analyze data using an AI-powered chat experience in canvas and model-driven apps (preview) in their environment. For more information, see Manage feature settings.

Set Copilot feature ON for the envrironment

When the feature setting is turned on, a maker can then enable Copilot component from the app settings in Power Apps Studio and then add the Copilot control to the app, allowing end users to use the AI-powered chat experience.

Step 2 - Enable Copilot component for a canvas app

Open your canvas app open for editing in Power Apps Studio:

  1. On the command bar, select Settings > Upcoming features.

  2. From the Preview tab, set the toggle for Copilot component to On.

    Turn on Copilot control.

Step 3 - Add Copilot control to your canvas app

Add the Copilot (preview) control to your canvas app enabling end users to gain insights about the data in their apps through the chat experience.

With your canvas app open for editing in Power Apps Studio:

  1. On the app authoring menu, select Insert and select Copilot (preview) to add this control.

    Add the copilot control.

Choose data for Copilot

  1. When the Copilot control is added to the canvas app, select a data source from the pane. Copilot can only provide data insights on a single Dataverse table when an end user asks a question.

    Select a data source.

    Or, from the control Properties tab, select Data source (Items) and choose a Dataverse table for your data source.

    Tip

    You can connect any Dataverse table in your environment as a data source for Copilot, regardless of whether the table is added to your canvas app or not.

  2. Select the specific Fields and/or View that the Copilot control answers questions for.

  3. To configure the selected table and its columns for use in Copilot, see Configure tables to use Copilot.

Note

Copilot can only answer questions for smaller datasets in a canvas app. The only exception to the dataset limit is when a Dataverse table is selected as the data source.

Provide feedback

App users and makers can provide feedback by selecting Like (thumbs up) or Dislike (thumbs down) button for each response that Copilot provides. Optionally, app users can also enter additional feedback in the text box and then select Submit.

Disable feedback for app users

Admins can disable the option for apps users to provide feedback to Microsoft from the Copilot chat experience.

  1. Sign in to Power Apps.
  2. On the left navigation pane, select Tables > Organization.
  3. In the Organization columns and data section, select the list of columns and search for Allow users to provide feedback for App Copilot.
  4. Set the toggle to No.

See also

Build apps through conversation (preview)

Add Chatbot control to a canvas app (preview)

Leverage Azure OpenAI Service in AI Builder (preview)

Add Copilot for app users in model-driven apps (preview)