The Expert's Curse: Why Your Best People Trust AI Least
The veteran with folded arms is not the last holdout. They are the smoke detector.
There is a moment repeating itself in offices everywhere right now, and almost nobody is reading it correctly. A team adopts a new AI tool. The junior staff love it instantly. They are faster, bolder, delighted. And the most experienced person in the room, the one everybody actually trusts, folds their arms and says, quietly, "I am not so sure about this." The usual interpretation is that the veteran is afraid, resistant, threatened by the shiny thing that makes a career's worth of hard-won skill look ordinary. The usual interpretation is almost exactly backwards.
What expertise actually is
Expertise, underneath everything, is the ability to see what is wrong. A beginner at chess sees a reasonable move. A grandmaster sees, instantly and without effort, that the reasonable move loses in nine. The grandmaster did not memorise more moves. They spent years building an instinct that flinches at the almost-right, the plausible, the thing that looks fine and is quietly fatal.
Now hand both of them the same AI. It produces fast, fluent, confident, polished work. The beginner sees magic. The expert sees the eight percent that is subtly off, because spotting the eight percent is the entire thing they spent twenty years learning to do.
The veteran's hesitation is not fear. It is detection. They are reading the tool better than anyone else in the room.
Enthusiasm and competence point in opposite directions
This creates a genuinely strange problem. The people most equipped to use AI well are the most reluctant to, and the people most eager to use it are the least able to catch its mistakes. The confidence in the room is inversely related to the competence.
The loudest adoption signal comes from exactly the people who cannot yet tell when the tool is wrong. The most reliable quality signal comes from the person quietly raising an eyebrow in the corner.
The skeptic is the smoke detector
The mistake leaders make is to treat the skeptics as obstacles to win over and the enthusiasts as proof the rollout is working. It is the reverse. Your enthusiasts show you adoption. Your skeptics show you the truth.
The veteran is not the last holdout. They are the smoke detector. When the person who can tell good from almost-good says wait, the right response is not to convince them. It is to listen.
None of this is an argument against AI. The experts are not saying no. They are saying not like this, not yet, not without me checking. That is not the sound of resistance. It is the sound of judgment, which is the one thing in short supply.
The fair objection: sometimes experts are just wrong
History is littered with experts who folded their arms and were simply protecting their turf. The scribes who distrusted the printing press, the craftsmen who smashed the looms, the editors who swore the internet was a fad. Expertise can curdle into self-interest, and "I am not so sure about this" is exactly what a threatened incumbent says too. So how do you tell detection from defensiveness?
Ask them to be specific. Real detection can point at the flaw: here is the case it will get wrong, here is where it breaks, here is the eight percent. Turf-guarding cannot. It can only repeat the feeling. A skeptic who names the exact failure is data. A skeptic who just says "I don't trust it" might only be defending the old map. Listen hardest to the ones who can show their work.
The line worth remembering
The future will not belong to the people who trusted AI the most, or the people who trusted it the least. It will belong to the ones who learned to trust it exactly as much as the experts do. Which is to say, warmly, constantly, and never completely.
I write about AI, data, and learning regularly at pinaldave.com, and I have been teaching this hands-on in my AI workshops.