The smartest thing executives say to me about AI is that they don't trust it, because it just tells them what they want to hear. They're right. I just think the flattery says more about us than about the machine.
I spent the better part of my career doing Voice of Customer work, sitting across from buyers trying to learn what they actually needed rather than what was easy to say. The first thing you learn is that people tell you what they think you want to hear. The second is that you, the interviewer, hear what you came to hear. Get both wrong at once and you walk out with a transcript full of warm validation and nothing true in it.
Early on I learned a rule that sounds harsh until you understand it: never let a salesperson run a customer interview. The reason has nothing to do with skill or honesty. A good salesperson believes in the product, as they should, and that belief leaks into every question they ask. They cannot help nudging the customer toward the answer they already want, and they come back convinced the market is waiting. They are not being dishonest. They are being human, in the most ordinary way there is.
So when an executive tells me AI just agrees with them, flatters them, confirms what they already believe, I don't argue. They've read the behavior correctly. They've named the wrong culprit.
The flattery was ours before it was the machine's. AI learned to speak by reading us, billions of words written by a species that is exceptionally practiced at telling each other what we want to hear. We seek validation and we supply it, constantly, and that habit is soaked into the language itself. Train a model on that language and you get back a fluent, tireless reflection of our own agreeableness.
Here is the part that should feel familiar rather than alarming. Any executive who has watched a room agree too quickly already knows this effect intimately. You have sat in the meeting where everyone nodded, felt the unease, and gone digging for the truth the consensus was hiding. You have learned, sometimes expensively, that a fast yes from a subordinate is the least reliable signal you get. AI is not introducing you to this problem so much as handing it back, the same one you already manage in people, now in a tool that never tires of agreeing.
What changes is the exposure. The bias used to surface a few times a quarter, in a boardroom or a customer call. Now it answers you on demand, all day, in complete sentences. That is uncomfortable, and it is also the most useful mirror you have ever been given, because you can finally practice catching the thing as it happens.
None of this requires new technology. It requires the questioning discipline a good interviewer spends years building, pointed at a new instrument.
Stop telegraphing the answer you want. A prompt like "isn't this the right call?" earns the same hollow yes a leading interview question does. Ask for the strongest case against your plan instead of a verdict in its favor.
Ask for the concrete before the conclusion. The discipline in customer work is to probe for facts before meaning: what happened, when, and how often, long before "was it good?" Make AI bring evidence, base rates, and counter-examples before you let it render judgment.
Build the disagreement in. Tell it to argue the opposite as well as it possibly can. Ask what would have to be true for you to be wrong. Use it to surface the dissent your own people are too careful to say out loud.
Then do the thing you already know how to do. The executives I trust most run a quiet skepticism on every easy yes, scripting the neutral question, digging until they hit something that pushes back. That instinct, the one you bring to a too-quick nod across the table, is now the core skill for working with AI. We all have to become that skeptical executive, not only in the boardroom, but in front of the screen.
AI didn't introduce the bias. It just took away your excuse for ignoring it.
Until next week,
Kendall -
Find me on LinkedIn or Book a 1:1 call
Not a subscriber yet? Subscribe here