A few months ago, during a call with a potential client, a sharp question landed on me—the kind that looks innocuous only until you have to answer.
“In your view, how long does this project take to close?”
I had everything I needed to improvise a credible answer. Years of experience, similar projects behind me, a gut estimate that could sound reasonable. Six months, maybe eight, with a prudent margin. Said in the right tone it would have looked the part: competent, reassuring, the voice of someone who knows their business.
Instead something else came out.
“I don’t know. Not yet. Give us two weeks to analyse the context and we’ll come back with a number we believe ourselves.”
The call went strangely silent. Three seconds, which in a sales conversation feels like a long time. Then the CEO on the other end said something that caught me off guard.
“Finally someone who isn’t pulling a number out of the air.”
We won that project. Not despite that “I don’t know”. Probably because of that “I don’t know”.
The age of instant answers
I often wonder when we decided that the speed of an answer was a proof of intelligence. Or competence. Or authority.
Today an answer is always available. About anything, at any time, in less time than it takes to phrase the question well. You search Google and in a fraction of a second you have endless results. You ask a language model and you get fluent text, well-written, even convincing.
And so, without noticing, we’ve internalised a slightly toxic equation: whoever answers fast knows. Whoever hesitates doesn’t. Whoever says “I don’t know” is out of the game.
You see it in meetings, where whoever speaks first often “takes” the room. You see it in job interviews, where hesitation gets read as unpreparedness, even when it’s just caution. You see it in consulting, where “let me think about it” feels like a loss of value, as if the client were paying for ready-made certainty in your pocket.
And then you see it in AI, which is maybe the most perfect embodiment of this mechanism.
A language model isn’t designed to stop and say “I don’t know”. It’s designed to produce an answer. Structured, plausible, maybe brilliant. Even when it doesn’t know. Sometimes especially when it doesn’t know. Hallucinations aren’t a folkloric incident, they’re the natural consequence of a system trained to always fill the void.
We’ve built a perfect machine for an age that decided silence is a defect.
The cost of fake certainty
The interesting question, though, isn’t why AI can’t say “I don’t know”. It’s why we struggle so hard to say it ourselves.
I’ve spent twenty years between technology and consulting. Meetings, calls, presentations, workshops. If I think back to the times someone openly said “I don’t know” without it sounding like a confession of guilt, I can count them on one hand. The unwritten rule is clear: you have to have an answer. Better wrong than absent.
The problem is that this norm has an enormous cost, only it’s hard to see while it’s happening.
You see it later, when a project estimated at three months drags on for twelve, and at some point someone says quietly: “we didn’t actually have enough information to estimate”. You see it when an architecture is chosen because someone spoke with great confidence in a meeting, and the honest version would have been: “let’s do a proof of concept before deciding”. You see it when a contract is born crooked, because the timeline got promised before anyone understood what was really underneath.
I’ve watched products fail not because competence was missing, but because the courage to admit uncertainty was. And the bitter part is that the people faking knowledge often get rewarded. Certainty, even fake certainty, is reassuring. It gives the illusion of control. And whoever offers it gets perceived as a leader.
It’s a slightly inverted incentive system: it rewards speed over truth, confidence over honesty, appearance over substance.
Three dangerous words
“I don’t know” is dangerous because it breaks a power dynamic.
In a meeting with a client, saying “I don’t know” means admitting that you, the expert, the one being paid to know, don’t know in that moment. It’s as if you cracked the implicit pact of consulting, where the client buys certainty and you sell it.
Inside a team, if you’re the lead, saying “I don’t know” is even more counter-intuitive. Because we’ve been taught that leading means having the direction, having the answer, having the plan. And no one wants to follow someone who admits not knowing the road.
Yet if I think back to the best decisions I’ve made, there’s almost always an “I don’t know” at the start. Not because ignorance is a virtue. It isn’t.
It’s because “I don’t know” is often the only honest starting point for a serious decision. It’s the moment you stop performing and start reasoning. It’s the move from the prepackaged answer to the real question.
When someone says “I don’t know”, odd things happen. The air lightens, as if it suddenly became permitted not to know. Someone finds the courage to say “I have a doubt too”. Information surfaces that had stayed hidden because no one wanted to look like the one slowing things down.
Maybe that’s the part that interests me most: “I don’t know” isn’t the opposite of competence. It’s often its prerequisite.
The AI paradox
Here comes the paradox.
AI makes it even harder to say “I don’t know”. Because if a machine produces an answer on any topic in two seconds, how do you justify that you, a professional with years of experience, don’t have one ready?
The benchmark shifts. The machine’s speed becomes the standard against which we measure human speed. And in that race, stopping to think becomes a luxury.
Except it’s the wrong benchmark, founded on a confusion: mistaking having an answer for having the right answer.
AI always has an answer. Often it’s good. Sometimes excellent. But it doesn’t know when it doesn’t know. And that, paradoxically, is one of the most precious capacities we have: recognising that a piece is missing, sensing that something doesn’t add up, intuiting that the variable being ignored is precisely the one that changes everything.
Socrates built a whole philosophy on it: knowing that you don’t know as the basis for actually knowing. Twenty-five hundred years later, we have machines that don’t know they don’t know, and we use them as a model of cognitive efficiency. The irony is almost perfect.
A quiet competitive advantage
My experience is what it is, limited and full of bias, like all experience. But it has left me with a fairly stable conviction: people and companies who can say “I don’t know” at the right moment win in the long run.
The clients I’ve worked with longest are often the ones with whom, at the start, I admitted I didn’t have all the answers. Not because they appreciated incompetence, but because that gesture created a different kind of relationship. A relationship where it was permitted to explore, ask questions, change one’s mind. Where the solution didn’t have to be ready by slide three.
The best people I’ve worked with are the ones who, in the middle of a meeting full of acronyms, said “I don’t understand” or “can you explain?”. The famous “stupid” question everyone has in their head and no one asks. Almost always it’s the one that unblocks everything.
And the best decisions, the ones that years later turned out to be right, had a moment of suspension before becoming a “yes” or a “no”. A small space of legitimate doubt. A refusal to answer just to take the pressure off.
It isn’t a method. It isn’t a framework. It isn’t a slide on vulnerability.
It’s something simpler and harder: accepting that competence includes recognising your own limits, and that those limits aren’t a shame. They’re the exact border where real thinking begins.
The last word
We live in an era that demands immediate answers about everything. Algorithms that respond in milliseconds, colleagues who expect answers in minutes, clients who want them in hours.
In this constant noise of cheap certainties, saying “I don’t know” almost becomes a subversive act. A small resistance against a system that mistakes speed for truth and confidence for competence.
I’m not making a case for ignorance. I’m making a case for the pause. For that uncomfortable moment between question and answer in which, if you resist the temptation to fill it with the first thing that comes to mind, something rare happens.
You actually think.
Key takeaways
Fake certainty is an inverted incentive system: it rewards speed over truth.
Your best clients are often the ones with whom, at the start, you admitted you didn’t have all the answers.
In an age of instant answers, saying “I don’t know” is a small subversive act.
Questions & answers
Why is saying "I don't know" a sign of competence, not weakness?
Because it distinguishes a grounded answer from an invented one. Saying “I don’t know, I have to check” does three things at once: it acknowledges the limits of your knowledge, signals to the client that you don’t improvise, and creates space for a reliable answer. People who give an answer to everything, on everything, are usually filling the air with words to avoid losing perceived authority—and getting the opposite effect over the medium term.
Isn't it risky to admit ignorance in front of a client?
It looks risky in the short term and is the opposite in the medium term. A client who hears you say “on this one I’m not sure, I’ll confirm tomorrow” learns to trust every other answer you give. A consultant who never says “I don’t know” eventually gets caught out on something specific—and all the accumulated trust collapses with it.
How do you cultivate this habit in a competitive context?
Three practices: (1) explicitly separate “I know” from “I think” from “I suspect” in your answers—the confidence level is useful information; (2) prepare transition phrases that don’t sound evasive (“good question, let me think for a second”); (3) recognise the pressure of not-knowing—it often comes from personal insecurity, not from actual reputational risk. The more you practise, the lighter it gets.
What does AI have to do with the value of saying "I don't know"?
LLMs have a structural incentive never to say “I don’t know”—they’re trained to always provide an answer, and they often hallucinate to complete the task. For people working with them, keeping the habit of saying “I don’t know” (and demanding it when needed) becomes a form of professional hygiene. The human who says “I don’t know” is the counter-habit the system lacks.