Microsoft has just released a preview of Request for Information (RFI) in Copilot Studio agent flows, and I think this is one of the most important updates yet for organizations building AI-powered automations.

Why? Because automation is powerful, but judgment still matters. In industries like insurance, finance, healthcare, or legal, there are moments where AI can’t (and shouldn’t) make the decision on its own. This is where human-in-the-loop becomes critical.

What RFI Does

With RFI, your Copilot Studio agent flow can now pause, collect structured input from the right person, and then resume automatically. This goes far beyond a simple “approve or reject.” You can design a request for very specific inputs: numbers, text, yes/no answers, even dates. That information then flows straight back into your automation.

Real-World Business Applications

I see huge value for organizations in scenarios like:

• HR recruitment → A candidate screening bot could pause to request a hiring manager’s assessment on culture fit before moving to the next stage.
• Field service → A maintenance workflow could stop until a technician provides on site condition details or confirms safety checks.
• Healthcare administration → A claims process could request a specialist’s review notes before moving forward.
• Public sector services → Citizen applications could pause while an officer validates supporting documents.
• Manufacturing quality checks → Automated reporting could hold until a quality inspector provides test results or signs off on exceptions.

Why It Matters for AI Ethics and Governance

This feature isn’t just convenient — it directly supports responsible AI principles:

• Accountability → Keeps humans in control of critical decisions.
• Transparency → Provides a clear record of when and why human input was required.
• Fairness & Reliability → Ensures complex, context-heavy judgments aren’t left to automation alone.

In other words, RFI helps strike the right balance between efficiency and human oversight, giving organizations confidence that their automations are not only fast, but also accurate, compliant, and ethical.

I would love to hear your thoughts.

👉 Have you experimented with RFI in Copilot Studio yet?

👉 What types of processes would you want to keep human judgment involved in?