The One Question to Ask Before You Let AI Do Anything
Part 5, the capstone, of a 5-part series on AI myths and misconceptions. Previously: Your AI Sounds Confident. That's Not the Same as Right.
Four myths, one thread running through all of them.
The cash machine didn't wipe out bank tellers. It freed them up for the work that needed a person (Bessen, IMF, 2015). The cheapest bot didn't save money. It quietly cost customers. The wall that keeps people away from your staff loses you the customers who needed help most. And the confident AI that sounds right is sometimes flatly wrong, with your business holding the bill.
Different myths. Same lesson underneath. And that lesson hands you one question to ask before you let AI touch any part of your business.
The question
Does this take the routine volume off my people so they can do the parts that need a human, or is it trying to replace the human judgment itself?
That's the whole test.
If a tool clears the repetitive work, the same questions, the after-hours calls, the follow-ups, the sorting and the drafting, so your people have room and time for the judgment, the empathy, and the hard calls, that's the good kind. It makes your team more valuable, not less. It's a partner doing the heavy, boring lifting so the people can do the human part.
If a tool is sold to you as a way to remove the person and hand their judgment to a machine, be skeptical. That's the version that looks cheap, walls off your customers, and states wrong answers with a straight face. Every myth in this series is a different face of that same mistake.
What a real partnership looks like
Think of it as a division of labor, not a swap.
The machine is good at volume, speed, and never getting tired of the same question. It is bad at judgment, and it has no stake in whether your customer stays or goes. You and your people are the opposite. Slower at volume, far better at the call that actually matters, and you are the ones who carry the relationship and the responsibility.
So you keep those. The AI takes the load off. You keep the judgment, the final say, and the customer who needs a human. That's not AI running your business. That's AI doing the parts of the work that don't need you, so you can spend your hours on the parts that do.
The businesses that win with this aren't the ones that hand over the most. They're the ones that stay clearest about where the machine ends and the person begins.
Frequently Asked Questions
Do I need to run this test on every tool, or just the big decisions?
Run it on anything that's actually speaking or deciding on your behalf, not on how much money happens to be at stake. A tool that only organizes your own notes stays fully inside your control. One that talks to customers, or acts before you've reviewed it, has stepped into the territory this question is actually about.
What if I genuinely can't tell which side of the question my situation falls on?
That uncertainty is useful information on its own. It usually means you haven't mapped out what the tool actually does in enough detail yet, not that the answer doesn't exist. Go back to the specific task, and the answer usually gets clearer fast.
A simple next step
Take the one place you're thinking about adding AI. Just one. Run it through the question at the top of this article.
Is it taking volume off your people, or taking over their judgment? Can your customer still reach a human when it counts? If it's wrong, who catches it, and what does that cost you?
If it would help, bring me that one decision and I'll run it through the test with you, out loud, together. No contract, no obligation. You'll leave knowing whether it's the kind of AI that makes your team stronger or the kind that quietly costs you, and that's worth knowing whether you ever work with me or not.
Sources & Further Reading
Toil and Technology, James Bessen, IMF Finance & Development, March 2015
What Bank Tellers and Radiologists Can Tell Us About Our Job Security in the AI Era, CCIA, 2026
