Introduction to Microsoft Copilot : Using Dataverse as a Knowledge Source in Microsoft Copilot Studio

 Hello Guys,

In our previous blog, we learnt how to publish our agent in Teams channel. Today we will see how to add knowledge in Agent.

Do follow us to know more about Dynamics and Power Platform. Also if you like our blog then please comment and share this blog with your friends.

This blog explains how Knowledge works in Microsoft Copilot Studio and how you can use it to make your agent more intelligent by providing it with business data.

In our previous blogs, we created Ravisha, our travel agent, which helps users plan trips and perform related travel activities. To make Ravisha more powerful, we can provide it with knowledge sources that it can use to answer user questions with accurate and relevant information.

For example, imagine we have a Trip table in Dataverse that stores travel-related information. If users want to ask questions about their trips through the Ravisha agent, we first need to provide the agent with access to this data. This data source becomes the agent's knowledge, allowing it to retrieve and present the required information.

Copilot Studio supports multiple knowledge sources, such as websites, documents, SharePoint, and Dataverse. In this blog, we will use Dataverse as our knowledge source.

Let's see how to add knowledge.

As shown above, we already have our Ravisha agent created. Now Navigate to the Knowledge tab.

Click Add Knowledge.

Select Dataverse as the knowledge source.

Search for the required table and select it.

You can see that the Trip table has now been added as a knowledge source.

Open the Trip knowledge source and click Synonyms. Here, you can define alternative names and values to help the agent better understand user queries and improve search accuracy.

Save your changes. Next, go to Topics and open System Topics. Click on Conversational boosting.

Select Datasources.

Under the Knowledge section, enable Search only selected sources. Then select the Trip knowledge source.

Disable the Allow the AI to use its own general knowledge option. 

This ensures that responses are generated only from the knowledge source that you have configured.

Click Save.

Publish the agent.

Now your Ravisha agent is ready to use the Dataverse Trip table as a knowledge source. You can test the agent in Microsoft Teams and ask questions related to trip information stored in Dataverse. The agent will retrieve and return answers based on the knowledge you have provided.

Hope it helps...