Customize Microsoft 365 Copilot Intelligence with Copilot Studio

In a world where AI technology is changing rapidly, Microsoft 365 Copilot As a powerful productivity tool, it is gradually changing the way people work. With Copilot Studio, users have the ability to customize even further. With Copilot Studio, you can easily customize Microsoft 365 Copilot with your own intelligent agents and extend Copilot's capabilities to a wider range of areas through the Actions feature, such as interacting with external databases like Azure Data Explorer.

In this article, we'll walk through a simple example that demonstrates how to leverage the Copilot Studio and Actions features to customize Microsoft 365 Copilot into an intelligent Agent that can interact with Azure Data Explorer databases.

Getting Started: Defining Action

First, the user needs to define an Action in Copilot Studio, which is the bridge between Copilot and the external system. The following steps show the flow of Action definition:
Create a new Action: In the Copilot Studio interface, the user can find the entry point to create an Action and start defining a new Action.

The

Using Actions in an Agent

Once the Action is defined, you can use the Action in an Agent created in Copilot Studio to connect the Agent to the Azure Data Explorer database, giving the Agent the ability to query the database.

1. Add an Action to the Agent: In the Agent's configuration screen, find the Add Action option and select the Azure Data Explorer Action you defined earlier.

   

2. Configure the details of the Action: Users can further configure Action details such as server address, database connection information, authentication method, etc. This information will be used for the Agent to connect to the Azure Data Explorer database at runtime.

   

3. Set the Prompt: In order for the Agent to better understand user intent and generate the correct database query, you can set up prompt words to guide the Agent's behavior. The following is a sample prompt that instructs the Agent to generate a Kusto query and execute it:

   

   - Generate Kusto queries based on user requests.
   - Execute the generated Kusto queries.
   - Return the results of the executed queries to the users.
   - Ensure the queries are accurate and efficient.
   - Provide clear and concise responses to users.
   - Handle errors gracefully and inform users of any issues.
   - Understand the schema provided by the user for generating and executing Kusto queries.
   - Table name is StormEvents.
   - The schema includes columns such as StartTime, EndTime, EpisodeId, EventId, State, EventType, InjuriesDirect, InjuriesIndirect, DeathsDirect, DeathsIndirect, DamageProperty, DamageCrops, Source, BeginLocation, EndLocation, BeginLat, BeginLon, EndLat, EndLon, EpisodeNarrative, EventNarrative, and StormSummary.
   - Communicate in a casual manner.
   - Format the result as a table.
   

Demonstration of effects and multi-language support

Once you have completed the configuration of the Agent, you can test to see how well the Agent integrates with the Azure Data Explorer database.

Configuration effects are shown: The following figure shows the test results after the Agent configuration is complete. As you can see, Copilot Studio is able to successfully invoke the defined Action, connect to the Azure Data Explorer database, and execute the query raised by the user.

Safety and security: To ensure security, Copilot requests user authorization before performing external operations. Users can choose "Always allow" or "Allow once". When used for the first time, users are also required to complete authentication to ensure that only users with database access rights can execute queries, effectively safeguarding data security.

Query results are presented: The following figure shows the results returned by Agent after executing the query. The results are presented in a tabular form, which is clear and easy to understand, making it convenient for users to view and analyze the data in the database.

Multi-language query support: What's even more surprising is that users can use any language they are familiar with (theoretically, almost all languages are supported) to make data query requests to the Agent, and Copilot automatically understands the user's intention and converts the request into the correct database query statement, and finally returns the results. No additional language processing is required, which greatly enhances the ease of use.

• Chinese query example:

  

• Example of a Japanese language query:

  

Extended Application Scenarios

In addition to simple data query, users can also integrate data query action with other tasks to build more complex intelligent agent application scenarios. For example, you can combine the query database results with document generation, report analysis and other functions to realize richer functions.

• Integrating Other Tasks Example 1:

  

• Integrating Other Tasks Example 2:

  

With the examples in this article, you can see that with the Copilot Studio and Actions features, users can quickly and easily customize powerful and intelligent agents for Microsoft 365 Copilot and connect them to external systems such as Azure Data Explorer, greatly expanding the application boundaries of Copilot. This greatly expands the boundaries of Copilot's applications and brings a smarter and more efficient work experience to both organizations and individual users.