Insights
Why Employees Ask AI Questions They Hide From Peers
A colleague told me recently that she had asked an AI system a question she would never have asked her own team. It wasn’t a sensitive question. It was a basic one — something about how a part of her company’s business actually worked, a thing she felt she should have understood three years ago. She typed it into a chat window instead of turning to the person sitting eight feet away who could have answered it in a sentence. I have done the same thing. I suspect most people reading this have. We tend to file this under convenience. I think it is something else. It is a measurement.
The social-psychological cost
There is a long research tradition on why people don’t ask. Studies of help-seeking at work find that employees avoid asking colleagues for two reasons: a social-psychological cost — the fear of appearing incompetent — and an economic one — the worry that the question follows you into a promotion conversation. Field experiments on internal knowledge platforms show both costs suppress asking, and that anonymity partly removes them. When the seeker’s name is hidden, the questions come out.
A study published earlier this year in Knowledge and Process Management gives the darker version of the same finding. The authors name a construct they call knowledge suppression: the conscious withholding of valuable, unsolicited information because of interpersonal and socio-emotional barriers. Not hoarding — the person is not protecting an advantage. They have something worth saying and they decide the interpersonal risk isn’t worth it. Three mechanisms drive it: fear of the recipient’s reaction, a wish to spare someone’s feelings, and self-protection. The authors’ conclusion is one I keep returning to: the intelligence is being lost not through missing technology or missing requests, but through unmanaged social filters.
The safest colleague most people have ever had
Now place a machine inside that system. The AI does not judge. It does not remember your question at review time. It has no opinion of you, no rank, no meeting you are also in. It is, in a strictly psychological sense, the safest colleague most people have ever had. Of course we ask it things we won’t ask each other. The cost of asking has dropped to zero on one channel and stayed high on every other one. Which means something useful is now visible. The gap between what people ask machines and what they ask each other is a rough measure of how much social cost your organization imposes on curiosity. Every question routed to AI that a colleague could have answered better is a small reading on that instrument.
I want to be careful here. Asking AI is often simply the right choice, and I am not arguing for friction. What I am arguing is that a question asked of a machine ends where it started. Nobody learns that this thing is unclear to competent people. No process gets fixed. The person who could have answered never discovers what their expertise looks like from outside. The answer arrives, and the organization learns nothing.
The quiet cost
That is the quiet cost. AI can make individuals informed while leaving the organization exactly as ignorant of itself as it was before. So the question I would put to any leadership team is not “how do we get people to use AI more?” It is closer to: what does the fact that our people prefer asking a machine tell us about what it costs to ask us? Contribution requires exposure. A question is the smallest possible act of visible not-knowing. If the machine is absorbing all of them, we have not solved the problem of ignorance. We have privatized it.