Copilot Studio: Add Human Review to AI Agents
Diesen Beitrag auf Deutsch lesen
Use Teams or email to collect a human decision and steer a Copilot Studio agent when a situation is unexpected or too risky for autonomous action.
Posted 
Governance card: Autonomous agents need human review when decisions are risky or unexpected.
By Power Platform Governance
1 min read
TL;DR
Copilot Studio can add a human-in-the-loop step to an agent: the agent sends a message through Teams or email, collects the person’s response, and uses that response to decide its next action when situations are unexpected or too risky for autonomous decisions.
Original by Matthew Devaney, on Matthew Devaney. Read the original
This is our own summary, not a republication or full translation.
Governance takeaway
- Makers: Add a human-in-the-loop step where an agent may face unexpected situations or decisions that are too risky to make alone.
- Security/Compliance: Watch for autonomous decisions that require human oversight, and use the Teams or email response step to keep a person involved.
- Leadership/Business: Identify decisions that agents must not make independently, because the source warns that planning cannot cover every situation.
This post is licensed under CC BY 4.0 by the author.