Copilot Studio: Agent Request for Info

Key insights

• Request for Information (RFI)
  RFI pauses an agent flow to ask humans for missing or uncertain data, then resumes automatically once answers arrive.
• Copilot Studio agent flows
  The RFI action is built into Copilot Studio and fits directly into AI-driven workflows without extra integration work.
• Human-in-the-loop
  Configure a title, message, assignees, and input fields; the flow sends a form to reviewers and waits for their responses.
• Structured input
  Request specific types of answers—text, numbers, dates, choices or yes/no—so the flow receives validated, ready-to-use data.
• Improved decision accuracy
  Adding human review reduces failures, handles edge cases, and gives better context for automated decisions.
• Best practices
  Assign multiple reviewers for coverage, test flows end-to-end, and validate inputs to keep processes fast and reliable.

Overview of the demo

The YouTube demo presented by Microsoft during a recent Microsoft 365 & Power Platform community call showcased the new Request for Information feature inside Copilot Studio agent flows. The presenter, Derah Onuorah, walked viewers through a practical scenario in which workflows pause to ask humans for structured input, then resume automatically once the data is validated. Consequently, the video frames this feature as a bridge between automated logic and human judgment, helping teams handle exceptions and missing data more reliably.
Moreover, the demo emphasized how these pauses surface only when needed, which reduces unnecessary interruptions while preserving oversight on critical decisions. In short, the video positioned the feature as an attempt to combine speed with accuracy in business automation.

How the RFI action works

In the demo, the Request for Information action halts a flow at a chosen point and sends a structured request to designated reviewers by email. The messages include user-friendly forms with tailored input types such as text, dates, numbers, dropdowns, or yes/no fields, and once reviewers respond the flow resumes with the captured values driving subsequent steps. This design uses familiar email pathways, specifically actionable messages, so recipients can reply without switching tools, which simplifies the reviewer experience.
Furthermore, presenters showed that multiple assignees can be added to a single request to create redundancy or distribute responsibility, and the flow uses the first valid response or a chosen aggregation method to proceed. Therefore, the mechanism supports both rapid responses and careful review depending on how teams configure assignments and inputs.

Benefits and practical tradeoffs

The demo highlighted clear benefits: improved decision accuracy, fewer flow failures due to missing data, and richer inputs than simple approval toggles. Because humans provide context-sensitive answers, the system can handle edge cases that automated rules might miss, and automated processes remain uninterrupted once complete. As a result, organizations can expect higher quality outputs from workflows that need occasional human expertise.
However, these advantages come with tradeoffs. Pausing flows introduces latency and dependence on human availability, which can slow processes under tight deadlines or with poorly chosen assignees. In addition, configuring detailed input types increases design complexity, and teams must balance the level of information requested with the reviewers’ time to avoid creating bottlenecks.

Implementation and integration considerations

The video demonstrated a straightforward setup inside Copilot Studio, where authors add the RFI action, define titles and messages, select assignees, and choose input fields. Because the RFI uses email actionable messages, it integrates naturally with standard Outlook workflows and requires minimal additional infrastructure, which lowers the bar for adoption. Consequently, many teams can pilot the feature quickly while leveraging existing collaboration practices.
Still, the demo also suggested practical cautions: designers must test flows thoroughly to ensure requests trigger at the right time and that responses map correctly into follow-up actions. Moreover, maintaining audit trails and ensuring the right people receive requests demands careful role design and governance, especially in regulated environments where data accuracy and compliance are critical.

Challenges and balancing governance

While the RFI action reduces outright failures by collecting missing data, it also shifts part of the automation burden onto people, so organizations must craft clear policies for when human input is required. For instance, teams should decide which decisions warrant an RFI versus those that can be handled by rules or a simple approval, and these choices affect speed and risk. Additionally, the need to keep requests concise yet complete forces tradeoffs between thoroughness and reviewer fatigue.
Security and privacy add another layer of complexity, because the system sends potentially sensitive details via email forms; therefore, teams must ensure proper access controls, data handling rules, and monitoring. In this way, the feature can improve outcomes only if organizations invest in governance, role clarity, and thoughtful flow design.

Getting started and expected impact

To begin, the video advised users to open Copilot Studio, add the Request for Information connector into an agent flow, and then test the flow end-to-end to confirm the request and resume logic. Since setup relies on built-in connectors and Outlook actionable messages, pilot projects can move quickly, allowing teams to validate benefits in a few cycles. Over time, teams can refine where RFIs appear and automate follow-ups for common responses to reduce repeated human touchpoints.
Looking ahead, the demo suggested that combining human insight with automation will become a standard pattern for workflows that must balance speed with contextual judgment. Nonetheless, realizing that potential will require careful design, continuous testing, and a clear strategy for when to involve humans so that automation remains both fast and reliable.

Conclusion

The Microsoft demo on YouTube presented a practical and measured introduction to the Request for Information action in Copilot Studio agent flows, showing how human input can be integrated without disrupting automation. While the feature promises better accuracy and fewer failures, the tradeoffs around latency, complexity, and governance deserve attention before wide rollout. Ultimately, the approach offers a useful compromise: automate what is routine and call in people when context matters, provided teams plan carefully and monitor results.
As the community experiment grows, organizations will need to adapt these patterns to their own processes and compliance needs, balancing speed, reliability, and the human judgment that machines alone cannot fully replicate.

Keywords

Copilot Studio request for information, Copilot Studio agent flows, request for information Copilot agent, RFI Copilot Studio tutorial, Copilot agent prompt design, Copilot Studio best practices, configuring agent flows Copilot, Copilot Studio RFI examples