Copilot Studio: RFI Agent Flows Guide

Key insights

• Request for Information (RFI): RFI is an action in Copilot Studio agent flows that pauses a workflow to collect missing or critical data from people, then automatically resumes the flow when responses arrive.
  
• Human-in-the-loop automation: RFI combines AI automation with human judgment to reduce errors, avoid manual restarts, and ensure decisions use complete, accurate information.
  
• Setup steps: Add the RFI action to an agent flow, set the title and message, choose assignees, and define required fields; the flow sends an actionable message and waits for form responses before continuing.
  
• Structured RFI forms: Forms support text, numbers, dropdowns, and multi-select fields, with options for required versus optional inputs and multi-reviewer assignment or first-response handling.
  
• Typical scenarios: Use RFI for visitor access requests, compliance checks, insurance claims, and legal reviews where human verification or additional details are needed during automation.
  
• Operational benefits: RFI improves data quality, speeds resolution, provides end-to-end visibility, and integrates with Copilot Studio features like AI prompts, monitoring, and Power Automate interop.
  

Overview

In a recent tutorial video by Reza Dorrani, Microsoft’s Copilot Studio gains new practical depth with the introduction of the Request for Information (RFI) action for agent flows. The author walks viewers through how RFI embeds human judgment into automated processes, enabling flows to pause and collect missing input before continuing. Consequently, teams can reduce failures from incomplete data and avoid manual restarts while keeping decision points visible and controlled.

Related resources: Microsoft 365 | Power Automate | Power Apps | Power BI

Moreover, the video demonstrates integration with email-based workflows, showing how reviewers receive actionable forms and send responses that resume the flow automatically. This mix of automation and human oversight aims to improve both data quality and resolution times. As a result, organizations can scale complex workflows while preserving necessary checkpoints for compliance and nuance.

How RFI Works in Agent Flows

The RFI action allows designers to insert a structured request into a flow that pauses execution until a human provides the requested information. Designers configure titles, messages, assignees, and input fields such as text, numbers, dropdowns, or multi-select choices, and then the flow waits for responses delivered through email-based actionable messaging. Once input arrives, the flow consumes the data and continues with downstream logic, which keeps processes both automated and adaptable.

Additionally, the system supports multi-reviewer assignments and options like first-response processing to enable flexible team workflows. This flexibility helps organizations route requests to the right experts and avoid bottlenecks while keeping audit trails intact. By using a visual designer or natural language prompts, business users can add RFIs without heavy developer involvement, streamlining adoption.

Demo and Real-World Use Cases

In the demo, Dorrani walks through a Visitor Access Request scenario where AI validates incoming details and triggers an RFI when information is missing or inconsistent. Reviewers receive an email with an actionable form, supply the missing data, and the automation resumes to complete access provisioning or escalate as needed. This concrete example shows how RFI reduces back-and-forth, improves accuracy, and speeds up outcomes in routine processes.

Beyond access requests, the RFI pattern applies to compliance checks, insurance claims, contract reviews, and other workflows that combine structured automation with expert judgment. For example, a claims workflow can pause to request assessor details, while a legal review flow can ask counsel for a specific clause interpretation before proceeding. Consequently, organizations can target RFIs to moments where human insight is most valuable, which both preserves throughput and limits risk.

Tradeoffs and Challenges

While RFI brings clear benefits, teams must balance automation gains against potential delays when waiting for human responses. Pausing flows introduces latency that, if not managed, can slow overall processing and frustrate users. Therefore, designers should weigh which points truly require human oversight and implement SLAs, reminders, or fallback paths to limit unintended slowdowns.

Moreover, adding human steps increases complexity in error handling, security, and auditing. Organizations must design input validation, role-based access, and robust retry logic so that RFIs do not become single points of failure. In addition, teams should plan for scale: multi-reviewer setups and high request volumes require clear assignment rules and monitoring to avoid bottlenecks and uneven workloads.

Adoption, Best Practices, and Next Steps

To adopt RFI effectively, start with high-impact scenarios where missing data routinely blocks progress and where human expertise clearly adds value. Pilot with a limited set of flows to tune form fields, assignment rules, and response handling while measuring latency and completion rates. Then, expand to other processes using lessons learned to create templates and guardrails that promote consistency.

Finally, organizations should combine RFI use with monitoring and analytics to spot recurring gaps in upstream data and to reduce future RFIs through upstream validation. Training reviewers on expectations and using reminders or escalation rules will also keep processes smooth. Overall, the tutorial by Reza Dorrani highlights a practical step toward resilient automation that balances AI speed with human judgment, while also making clear the operational tradeoffs teams must manage as they scale.

Keywords

Copilot Studio agent flows, RFI action tutorial, Copilot Studio RFI workflow, Agent flows tutorial for Copilot, Request for Information workflow Copilot, Microsoft Copilot agent examples, Build RFI actions Copilot Studio, AI agent flow tutorial