Home / News / Copilot in model-driven Power Apps
17 November 2023
By: Vincent Pannenborg

Copilot in model-driven Power Apps

An AI assistant that can query the data inside your model-driven App by asking questions.

Introducing Copilot for Model-Driven Power Apps

Microsoft Copilot for Model-Driven Power Apps is an AI assistant that allows users to query data inside the app, making it easy to find information that otherwise is only possible by querying or creating reports. Copilot can turn any Model-Driven App into an intelligent app which could be the default apps like Sales or Customer Service Hub, or a custom App.

It offers a brand-new conversational experience where users can easily ask questions about the data in their app and quickly receive responses in natural language. Copilot can even help users navigate their app. At the time of writing, it’s in Preview and only availeble in US region environments.

Setup Copilot in 2 steps

The first step is to enable Copilot on an environmental level. To do this, you need to have Administrator rights within the specific environment where you want to enable Copilot. From the Power Platform admin center, go to Environments > open the environment you want > Settings > Features. You need to set 2 options.

Behavior Settings. Set the Model-driven apps release channel to “Monthly channel”.

Features. Enable the second Copilot option.
The second step is to enable Copilot on a table and column level. By default, Copilot does not search everything. You control what data Copilot can access by enabling tables and columns for search, which will be included in Copilot search and responses. By default, most standard tables are enabled for Copilot.

For Copilot to use a custom table, the “Track changes” and “Appear in search results table” table options must be enabled. Depending on the size of the data in the table it might take between fifteen minutes and a full day before all data will be available for questions.

To configure your columns for Copilot, open the table, and then select Views from the table hub. In the list of views, open the Quick Find view for the table. Add columns to the list in the “Find by…” options. Make sure to Save and publish your changes.

The Copilot experience

On to the fun part. Copilot can be accessed through the Copilot icon in the right navigation bar when you play a model-driven app. The pane can be opened or minimized as desired. Copilot can answers questions about data in the app, and it can also help you navigate the app. To demonstrate the power of Copilot, I use the Sales Hub and ask questions about sales opportunities.

You can ask questions that fall into four categories:

  1. Search for data where you would normally build a filter
  2. Insights that are otherwise only possible with charts or reports
  3. Special commands
    4. Navigate the system

Let’s start with questions that typically fall into the first category, questions that now require you to build a filter, use the column filter and/or sort data.

The answer provides the names of the opportunities, but there’s not much you can do with it yet. Alternatively, you can start your question with “List”. Copilot then gives you the option to navigate to a page with a view of the relevant Opportunities.
Examples of questions that fall into the second category are questions that require calculating values ​​and/or grouping data. These questions are really hard to answer unless you spent time building reports.
Copilot assumes that you want to see the average value here based on the “estimated revenue” of open sales opportunities. There are two revenue fields. So to get the value of opportunities won, you need to be more specific.
Now let’s take it a step further and really put Copilot to the test.
An example of a special command is Summarize.
Copilot can also help you navigate the app.
This use of AI to query data is a truly exciting development. Not everyone is an expert at building filters, let alone building reports. With Copilot, you can find the information you want by simply asking the right questions.

Happy Asking!