The most useful way I’ve found to manage AI isn’t as a tool — it’s as a brilliant new hire, still learning your world. That reframe turns AI from a tech problem into a management problem you already know how to solve.
Capable, but new to your world.
It’s extraordinarily capable, but it doesn’t know your seamer, your process, or your definition of “good.” Like any new hire, that’s on you to teach — you onboard it with structured data, and it learns your world.
The one alien trait: it never says “I’m not sure.”
A human new hire signals when they’re guessing. AI hands you a wrong answer with the exact same confidence as a right one. That’s what makes it more dangerous than a green employee — and why you can’t just ship what it gives you.
So you supervise it against a source of truth.
You measure its output against something you know is true — the way you’d check a part against a spec — until you’ve earned trust in it. That quality-control instinct is exactly the right one.
So what, now what.
So what: AI isn’t a magic box; it’s a brilliant, confident, context-blind new hire. Now what: before you rely on an AI output, ask “what’s my source of truth for checking this?” If you don’t have one, you’re not ready to trust it yet.
Leave a comment