By industry.
Industries differ mostly on the second question — what it costs you the day AI gets it wrong. The first question travels everywhere. But the tier at which a failure stops being fixable moves sharply from sector to sector, and that is what decides which legs you can assign and which calls have your name on them. Click any Dig deeper for where to start and where the hard line sits.
Five sectors
The two questions are the same across all of them. What it costs you to be wrong is not.
Financial Services
Which work costs you something you can't get back when it's wrong?
In financial services the second question does most of the work. Internal analysis and reporting cost you a redo. Fraud signals, compliance review, and credit decisions cost you something that doesn't come back.
Legal
Is a human reading every output before it creates an obligation?
Legal documents have a clean definition of done, which AI is good at. But a wrong clause creates an obligation before anyone catches it — so most of legal lives at Review, and the review has to be real.
Healthcare
If this is wrong, does a patient receive worse care?
Healthcare has the most absolute version of the second question. The administrative and operational load is enormous and assignable. Clinical judgment is a checkpoint, always.
Professional Services
Are your senior people the bottleneck — and is it because they're doing work anyone could do?
Senior people buried in assembly work, freed, and redirected to the work only they can do. That is the whole shape of it in professional services.
Industrial & Manufacturing
Where does an error reach the plant floor before a human can catch it?
Small percentages compounding across a large organization, one function at a time. The stakes swing harder here than in most sectors — and at the top end they include people.
Want this drawn for your sector specifically?
Bring your Map — a rough one is fine. Rough is how they all start.
From Running at the Speed of Light, Chapter Seven · Prometheus Press