I have created a copilot that collects information from users for a qualitative research project. I'd now like to analyse all the answers (variables) together and use an AI plugin to summarise the content in either word or in an app. I'm now sure how to connect the copilot data and run the AI summary. How do I do this?
Thanks will give this a try.
Not sure what "real time" means in this context. If you need to summarize responses or extract other insights post-conversation, you could access the transcripts Dataverse table and use an LLM (e.g. Azure OpenAI) to process the data.
It somehow works... in the `Custom Data` property, add a table of records such as
[ { Content : "blabla my data" } ]
in the `Input` property, add a search term query (not question...) which is sure to hit this data. And finally, in the `Custom Instructions` property, tell how you want this custom data to be summarized.
https://learn.microsoft.com/en-us/microsoft-copilot-studio/nlu-generative-answers-custom-data
This is worth exploring when sharing content with the user. Do you mean export variables to create custom data then inject back into generative answers later in the conversation to summarise?
This would be outside the scope of the co-pilot. So in real time our team could access a summary of all the answers provided from interactions with the copilot.
What the OP user means is that there is an undocumented confusion as to what the "Generative Answers" node do. You should perhaps rename it to "Search then Generate AI answers from the Search results" node. So it is not really a ChatGPT-like chat completion API as the OP user is expecting, it is used to search. Perhaps OP user can use the Custom Data property of the Generative Answers node to firstly inject the content that OP user wants to summarize, but I haven't tried this myself.
Do you need the summary to be "served" to users as part of the chat experience? Or is this outside the scope of the copilot?
Ekta Gupta
4
ricardodesouza
2
L-1234567-0
2