Risk Zones
A traffic-light filter for where it is safe to start with AI.
Green
Internal, text-heavy, verifiable, low-consequence. Start here.
Orange
Higher stakes but still internal - decisions, forecasts, calculations. More verification.
Red
Customer-facing or safety-critical. Not a first pilot.
Not sure?
Ask two questions: who sees the output, and what happens if it is wrong?
The 2am question
If this went wrong overnight and nobody spotted it until morning, how bad would it be? If the answer is 'not much', it's a safe place to start. If you wince, it needs someone in charge of it and a check before it acts.
When to use The first step in choosing a pilot - filter first, score second.
Read the full post →