Turn an ambiguous AI workflow into a safe, testable operating model. I review the proposed agent workflow and make the decision boundary explicit. You receive: A map of actors, tools, permissions, states, and human approval points. A risk and rollback register. A prioritized next-slice recommendation. A short handoff note your team can use. Best for teams evaluating an agent, API, or automation workflow before a large build. I do not promise autonomous operation or unverified business outcomes. We define what the system should decide, what evidence it must show, and where a person remains in control.