As the drive to embrace AI-powered automation in complex workflows continues apace, it’s remains clear that many critical business processes are likely to always require human input and analysis: supply chain quality control, financial services compliance, and IT security incident response, to name just three.
Microsoft is responding to this need for intelligence from humans in the loop by adding a new capability called Request for Information, or RFI, into Copilot Studio. RFI is designed to ensure that when human judgment is needed to keep workflows advancing, there’s a built-in mechanism to do so, and agents are smart enough to request that input automatically.
RFI is part of Copilot Studio agent flows, which automate repetitive tasks while integrating apps and services. Agent flows can be triggered manually or by other automated events or agents. Agent flows provide for simple workflow creation, consistent execution, end-to-end process visibility, and now, human-in-the-loop inputs.
Requesting Human Intelligence
RFI lets a workflow that’s in process pause execution to collect structured input from designated human reviewers before proceeding with subsequent steps. It then uses input from those human reviews in future steps.
Reviewers submit input via email-linked forms in Outlook and once they’ve done so, the flow resumes.
Microsoft outlines a four-step process for creating an agent flow with RFI functionality and a 10-step process for adding human analysis via RFI to an existing agent flow in Copilot Studio in this in-depth RFI blog post.
Some requests for input can be identified as optional; there are five input types are supported: text, yes/no, email, number, and date.
Microsoft noted in its RFI documentation that the first person’s response is utilized in the agent flow, and any subsequent responses that take place will not be processed. While requests are currently sent in Outlook only, more email platforms may be added in the future, the company said; requests can’t be sent to users outside of a company’s tenant.
AI Agent & Copilot Summit is an AI-first event to define opportunities, impact, and outcomes with Microsoft Copilot and agents. Building on its 2025 success, the 2026 event takes place March 17-19 in San Diego. Get more details.
Where RFI Applies
Referring back to some of the expected RFI use cases cited above, here is how the function would operate:
- Supply chain quality control: Manufacturing process automations can request inspection reports from quality assurance experts when automated systems detect potential defects or when supplier certifications need human verification.
- Financial services compliance: During loan applications or account reviews, workflows can pause to request more documents from compliance officers. These may include income verification, credit assessments, or regulatory approvals, to ensure the required level of documentation and analysis has occurred.
- IT security incident response: Security workflows can automatically request more details from cybersecurity analysts when threat detection systems identify suspicious activities requiring expert analysis that only human experts are able to deliver.
Each of these examples, and there are many others, clearly demonstrate that critical business processes today — and most likely long into the future — cannot operate without the expertise and oversight of humans with years of experience and direct expertise into the processes, policies, and mindset of the organizations they work in.
While fears of eliminating human roles crop up regularly in discussions around AI, agents, and automation, the human focus of this RFI functionality from Microsoft reflects a clear understanding that these powerful technologies can manage many functions — including requesting human knowledge and insight — but they’re no replacement for that same knowledge and insight.
Ask Cloud Wars AI Agent about this analysis
