What to Write Down When AI Gets It Wrong

Your AI system will make a mistake. That's not a scandal — it's a certainty. What separates responsible teams from the rest isn't avoiding mistakes. It's writing them down.

Keep an incident log. It doesn't need to be fancy. For every meaningful error, record five things:

1. When it happened — the date and time.
2. What happened — what the system did, in plain language.
3. Who was affected — a customer, an applicant, a patient.
4. What you did — who reviewed it, what the outcome was.
5. What changed — the fix, the new guardrail, the lesson.

That's it. Five lines. This log does three jobs at once: it helps you spot patterns, it proves to auditors that you take errors seriously, and it protects you — because "here's what went wrong and here's what we did about it" is the most trustworthy sentence in compliance.

The teams that get in trouble aren't the ones whose AI made mistakes. They're the ones who can't show they noticed, cared, and improved. Write it down.

I explain frameworks; I don't give legal counsel.