I wrote recently about executives who claim big gains from AI they cannot name. Now let’s talk about another person in the same room. The one who tried AI once, got a weak answer, and now tells everyone the whole thing is oversold. It makes things up. It is not that good. Case closed.
It took me a beat to see that this person and the overclaiming executive are actually making the same mistake.
Both skipped the work.
One skipped it and took the credit. The other skipped it and blamed the tool.
When someone tells me AI gave them nonsense, I would ask the same questions I would ask if a new hire handed me poor work. What did you actually request? Did you give it the context a capable person would have needed, or did you assume it could read your mind? What information was it working from, and was that information any good? When it asked you to clarify something, did you answer, or did you treat the question as a nuisance and push past it?
Most of the time the bad answer traces back to one of those, not to the tool.
The skeptics are not entirely wrong, to be fair to them. There is a version of the back and forth that feels like filler. The tool asks another question, then another, then suggests steps you did not ask for, and you start to wonder whether it is built to keep you talking and keep the meter running. My more cynical self wonders the same thing. It is a fair suspicion.
But most of the time, the questions are the work. Not something blocking the work – but the work itself.
The questions are the work. Not something blocking the work, but the work itself.
A good analyst does not take a one-line request and vanish to build the whole thing. A good analyst asks what you actually mean, tests your assumptions, and surfaces the details you forgot to mention. When a tool does the same, and you treat that as an obstacle rather than the job being done properly, you will always get a worse result. With no right to then blame the thing that was trying to get it right.
Both mistakes have the same root cause and it’s not the technology. It is an unwillingness to do the unglamorous part: giving good instructions and checking the work that comes back. The executive skips the measuring and reports a win. The skeptic skips the setup and reports a loss. Neither did the work of good data, clear context, and honest review. They just landed on opposite headlines.
Attacking the hype passes for wisdom. Doubt looks like rigor.
Notice which one earns more respect in a room. The skeptic sounds sharp. Attacking the hype passes for wisdom. Doubt looks like rigor. But dismissing a tool you never learned to use is not rigor. It is the same shortcut the executive took, better dressed.
I am not claiming AI is always right. Sometimes the tools genuinely fail. Sometimes the software behind it is weak. Those cases are real and a person doing this work should be able to tell the difference between a tool that failed and a task that was set up to fail. You earn the right to make that call after you have done the setup, not before. Before that, you are just guessing, the same as the executive who could not name that single specific gain.
So before you decide AI works or doesn’t, one question is worth answering honestly:
Did you do the work to find out? Or did you give it a lazy request, get a lazy answer, and take it as proof of what you already believed?
Both sides of this argument are louder than they have earned. The quieter people in the middle, the ones asking better questions and checking what comes back, are getting results that neither camp is talking about.
