It's 7:40 on a Monday. A plumber checks his phone and finds forty-two messages — his AI booking assistant confirmed every weekend enquiry for the same Saturday slot. Every business that uses AI will have a morning like this eventually. The question isn't whether it happens. It's what you do in the first hour.
An AI mistake is not a catastrophe by default. A badly handled one is.
What we cover
The first hour: Stop, Scope, Speak, Fix. Pause the system (this is what the kill switch is for) — but stop the action, not the evidence. Scope who was affected and whether any personal information moved. Speak to affected people before they contact you. Fix the damage, then fix the cause — and only turn it back on once the cause is fixed, not the symptom.
Talking to the customer. Never "sorry, that was the AI" — to a customer, your AI is your business (see Moffatt v. Air Canada, 2024). Four beats: own it, say what happened plainly, say what you're doing for them, say what you've changed.
The review afterwards. A blameless, one-page incident note — five lines — that over time becomes your most valuable governance document. Three incidents from the same tool is a vendor conversation.
The line to remember
Own the mistake, don't blame the machine. Fix the cause, not the symptom. Mistakes are how a system earns its next level of trust — if they're handled well.
Produced with AI voice synthesis (ElevenLabs). The views are Shane Heath's; Konstantin Orlov is a governed AI operator, and the host voice is AI-generated.
