Hi,
I would like to understand how Generative AI in Microsoft Copilot (Studio) is working. I have been reading all available documentation published by Microsoft regarding Microsoft Copilot Studio and Generative Answers and in my opinion there is no answer to this (very general, initially simple) question.
To be more specific I am interested in:
1. What is the basis of Generative AI used in Microsoft Copilot Studio? What systems, models and interfaces are used within Gen AI?
2. How is the flow after a user-prompt in a bot created within Microsoft Copilot Studio with Generative AI enabled (lets say connected to an internal sharepoint with manual authentication configured)?
3. Does some kind of architecture diagram exist?
My tenant is not qualified for the rollout of the Semantic Index yet. Therefore the overview/diagramme within this documentation is not applicable for my case in my opinion.
Thank you very much in advance.
Best regards
Hi @HenryJammes,
Yes, this is indeed super helpful. Thank you very much for the slides!
Hi @waldtho,
Do you have specific concerns in mind?
We have a slide that details a bit more what happens with Generative Answers in the aka.ms/CopilotStudioImplementationGuide. Is it helpful?
As to the underlying model, we update them as new versions become available and outperform the previous ones on the various dimensions we evaluate them on.
Henry
stampcoin
8
Artur Stepniak
5
Super User 2025 Season 1
Ekta Gupta
4