Create an effective autonomous agent

(1) Share

Report

Posted on by CU15090955-0

Hello. I'm currently developing an agent in Copilot Studio. I want my agent to be able to send emails, create Word documents, or create Word documents and email them, all depending on the user. Users have different behaviors. Sometimes they ask the agent directly and have the agent email them the answer, or create the answer in Word format. But they also often ask the agent first, the agent answers, and then they ask the agent to create the previous response in Word format or send the previous response to email. How do I address all these possibilities?

Categories:

Autonomous agents

I have the same question (0)

All responses (3)

Answers (0)

Suggested answer

Romain The Low-Code... 2,562 Super User 2026 Season 1 on at

Like (0)

Report
Copy link

Link copied!

hello, document generation come with the feature "code interpreter" you could use with 2 different way

inside the agent (in setting enable code interpret) or if you want more control from AI builder prompt (tool prompt from the agent)

here is the official documentation : https://learn.microsoft.com/en-us/microsoft-copilot-studio/code-interpreter-for-prompts

Was this reply helpful? Yes No
Michael E. Gernaey 53,960 Moderator on at

Like (0)

Report
Copy link

Link copied!

Hello @CU15090955-0

What you appear to be asking is, "how do I support them asking for documents before or after a question", not specifically (per say) how to build the document. I mean you may want that also, but you are really asking, how do I support providing the document at the Requesters leisure, not at some forced time

AKA, conditional document transfer to them based on them asking for it either initially within their question or after they asked the question, but didn't ask for them to download it.

Correct?

Was this reply helpful? Yes No
Michael E. Gernaey 53,960 Moderator on at

Like (0)

Report
Copy link

Link copied!

Either way, you are going to need to Capture the inputs and outputs and have conditional branches to track what they asked for and how, and then react too it.

You will always have access to the previous question if you store it in a Variable, so that essentially you are keeping state. Then you would have to decide... hey do I have multiple Topics, that get triggered based on their Request, or do you keep state in a single Topic, and as things come in you capture what they ask, you look at your variables in a condition or use a Generate / Create Node based on their question, where your data you have been capturing can be used as a knowledge source, including the whole conversation, allowing the agent to see what they want.

We would have to sit down and map out the scenarios fully to build it out correctly but the basis is above for what you need to do.

Was this reply helpful? Yes No