Before promoting a multi-agent mortgage underwriting workflow, compliance requires human review of a statistically valid sample of agent decisions, including tool inputs and final rationales. How should the team design this process in Microsoft Foundry?
Select an answer to reveal the explanation.
Short Explanation
You need a human review workflow with sampling criteria, structured rubrics, captured tool inputs and rationales, reviewer agreement tracking, and a release gate that blocks promotion until review service-level agreements and pass rates are met. That gives you auditable evidence in Foundry before you promote the workflow.
Full Explanation
Planning a human review process to evaluate multi-agent solutions in Foundry is an explicit exam skill. The correct design defines a human review workflow with sampling criteria, structured rubrics, capture of tool inputs and rationales, reviewer agreement tracking, and promotion gates that block release until review SLAs and pass rates are met, meeting compliance and quality needs. Shipping without sampling and relying on customer complaints is reactive and insufficient for regulated domains. Offline transcript skimming by one engineer with no rubric, sampling plan, or gate integration lacks methodology and gate integration. Automating every approval with a second agent and no human oversight removes required human oversight for high-stakes regulated decisions. Combine human review with automated evaluations so machines handle volume while humans validate residual risk.