I personally think that AI Skills fills the gap perfectly when looking to answer those Ad-hoc questions which cannot be catered for in the Power BI Semantic Models.

For example, while it is possible to calculate returning customers in Power BI, the measure is complex and use a lot of resources to get the result.

When comparing this to running the query using TSQL, the query is less resource intensive and can return the results quite quickly.

In this blog post I am going to show you how I combined the AI Skills in Microsoft Fabric, Copilot studio and Microsoft Teams to ask a question and get the answer within a few seconds!

The best part is that I am not going to use any code or even low code to achieve this.

AI Skill

The first part is creating the AI Skill.

I am not going to go into all the details on how to do this you can find more details here:

What I will say is that to use the AI Skill you need to be running an F64 or higher.

As shown below, when I am running my AI Skill I ask it the following question:

“How many customers after 2010”

And as you can see this gives me the answer I am looking for.

I now have got my AI Skill configured and ready to be used in the next steps.

Copilot Studio

The Copilot studio was new to me, and I did sign up with a trial.

Once I was signed in I then completed the following steps below to connect Copilot Studio to my AI Skill.

• I clicked on Agents and then “New agent”
• 
• On the next screen I clicked on “Skip to configure”, because I wanted to configure the agent.
• 
• On the configuration screen I put in the following details
• 

1. I renamed the agent to something more meaningful.

2. I also put more details in the description for other users.

The final step was to click on “Create”

• • • NOTE: We will add the AI Skills in the steps below. We first need to create the Agent.
• Once the agent is created, click on “Add knowledge” as shown below.
• 
• On the Add knowledge screen complete the following steps.
• 
  • 1. Click on “Advanced”
  • 2. Click on ”Microsoft Fabric”
• On the “Add AI Skills from Microsoft Fabric” screen I selected my AI Skill as shown below.
• 
  • I then clicked on “Add”
  • NOTE: It is possible to search for your AI Skill (As I am sure over time there will be quite a few AI Skills)
• Once completed I could now see it in the Knowledge section as shown below.
• 
• I then clicked on “Publish”
  • 
• I was prompted to click on “Publish” again due to the warning below.
  • 
• Once completed I could then see my Copilot Agent as shown below in the Agents section.
• 

Adding the Copilot Agent to Microsoft Teams

The final step is to add the Copilot Agent to Microsoft Teams and test the AI Skill.

• I open the Agent.
• 

1. I then click on the +4 to expand the other items.

2. I click on Channels as this is where I will find the channel for Microsoft Teams.

• On the channels screen I complete the following steps.
• 

1. The warning is because I have set the authentication to verify the users’ details when connecting.

2. I click on “Microsoft Teams”

• I click on “Turn on Teams”
• 
• Once completed I can see at the top that it was successfully added
• 

Adding the Agent to Microsoft Teams

• The easiest way I found to add the Agent to Microsoft Teams is to do the following
• NOTE: Make sure that your Teams is signed into the same tenant where the Agent was created.
• When I am in the AI Agent Chanel I go into the Microsoft Teams and click on “Availability Options”
• 
• I then click on “Copy Link”
• 
• I then take this link and open it in my browser that is signed into my Teams Tenant.
• When the link opens, I click on “Launch it now”
• 
• The Teams app opens, and I am shown the following below.
• 
• I click on Add
• Once added I get confirmation
• 

Asking Questions in Microsoft Teams

The final step is where I then started to ask some questions using the App in Microsoft Teams.

As shown below you can see that I asked a few questions which it responded to very well.

It was very quick, and I could keep on asking questions.

Summary

In this blog post I have shown you how I used a combination of new features in the Microsoft ecosystem to quickly and easily build an AI agent which can be used to answer questions from data.

This blog post and associated work did take me a while to get right, and I am certain that there might be other permissions and access required for this to work on a larger implementation. The idea of this blog post is to show you the functionality and how easy it has become to stitch these things together!

Thanks for reading!