In the last post I explained how you can build a CoPilot using the Azure OpenAI Studio in this blog post we will do the same with the CoPilot Studio. Here we will also utilize the Azure OpenAI Service. This is similar to the approach I used for my Intune AI Voice Bot, where AI is also used to interact with Microsoft endpoint management scenarios. Let’s have a quick look into this topic as well:
What are the prerequisites
- A license for Microsoft Copilot Studio (or an existing Power Virtual Agents license).
- It is also possible to start with an free trial
- Your Microsoft 365 tenant admin to deploy the Dynamics 365 and Copilot Studio app in the Microsoft 365 admin center.
- Access to the limited preview, by submitting a support ticket.
How to start
Open the CoPilot Studio via https://web.powerva.microsoft.com/ and login.
- Click on New copilot
- Enter a Name for your CoPilot and select the language.
Once this is done the bot is created and we have to add your custom data via skills. But before, let’s have a look at which functionalities are available
The Studio
There are basically 3 essential parts. #1 is the menu element, where you can select the options to create plugins, add your documents, build flows or everything else. #2 is the Copilot playground to test stuff and and validate that everything works after you configure it. The third part is the Main Part where you can take the configurations for the different options in #1.
Add your Documents
Go to Settings -> Generative AI and upload a document. In my case I used an export of Microsoft Learn docs.
That’s it? Let’s test it with asking a question about this document.
The answer is bad. Let’s have a look at what is missing. It was too easy. I forgot to activate the “Improve the reach of the conversation with generative responses“. Now the CoPilot answers with content from the documents.
Create an AI Plug-In
We have to create an PlugIn for this document. But how does it work? There are two kinds of PlugIns. One is an AI Plug-In like this:
Here you have an Input / Input Variables, and with this an Answer will be generated using GPT. This is an example of a Text classifier. You give the model a Text and some Categories, and it will assign a category to the text.
The second one is a Conversational PlugIn. Here you can build a kind of flow:
This flow offers you the full flexibility to connect to your data via the GPT Playground, make Conditions, Prompt for Auth and much more.
How to Integrate in Teams?
Also this is very easy. You can select Channels in the Settings and select Microsoft Teams.
If you want to compare this with the Azure OpenAI Studio way, check out how to create your own Intune Co Pilot using Azure OpenAi Studio.
- Select activate Teams and that’s it.
Now you will find your CoPilot in the Teams AddOns. If you enjoyed this, you might also like this post about how to build your custom GPT apps. If you want to go deeper into the concepts behind this, check out my Deep Dive into Co-Pilots.