Which one workflow are we inspecting, what outcome does it exist to produce, and which role owns that outcome?
What repeatable responsibility moved, and what evidence shows that another qualified person or system can restart, run, and check it without the original owner?
Which condition reads weakest on the evidence, what rival explanation remains open, and what next artifact would raise our confidence?
For the AI deployment: what may it not do, who notices a wrong output, who can change it, and where do exceptions go?
What verified harm, illegality, safety risk, or protected-worker impact would stop the work before optimization continues?
Which decision or handoff changed, what evidence now moves with the work, and who has the authority to act on it?
What credible next responsibility did the person receive, with what authority, learning, support, and credit?
Do these systems create opportunity after transfer while keeping Blank Collar instruments out of individual decisions?
Could the leadership role run through a planned absence, with direction, context, permissions, challenge, and accountability intact?
On what date will we review the evidence and decide to continue, revise, or stop?

One workflow. One evidenced constraint. One next artifact. One authorized owner. One date.

Evidence states: Works. Wobbles. Zero. No reading. The AI question alone may also return No deployment. A probe is not assurance.

Workflow economics: baseline outcome, full run and transition cost, error or risk measure, review date, and the decision to continue, revise, or stop.