You Can Tell When the Hand Is Off the Wheel

You can always tell.

The college student writing a paper on a subject they do not actually understand. The strategy document that reads like a consultant's template and says nothing concrete. The interview candidate giving polished answers that collapse the moment you ask one follow-up question. AI does not hide incompetence. It amplifies the appearance of it just enough to make the gap obvious to anyone who knows what they are looking at.

Where the hand off the wheel shows up

The most visible version is the student paper. The distinction worth naming is that AI use itself is not the problem. A student who uses AI to research a topic, reads the source material, actually learns what they are writing about, and then has AI help draft is using the tool well. The hand-off-the-wheel version is the student who asked AI to write a paper on a subject they did not bother to understand. The output is structurally correct prose the writer cannot defend. The teacher reads it and immediately knows. The format is right. The substance is missing.

The same pattern shows up in business. It is just dressed in better clothes.

In hiring, it shows up in interviews. A candidate gives a polished, articulate answer to a question. You ask one follow-up, and the answer collapses. Or you begin to suspect the candidate is reading from AI-generated suggestions in real time. The cadence is off. The pauses are too clean. The answers are too well-formed for questions they had no way to prepare for.

In documentation, it shows up as strategy decks and proposals that read like consultant templates. Generic frameworks. No concrete decisions. No reference to the organization's actual constraints. The kind of document that could have been written for anyone, about anything. Polished, professional, and entirely empty.

The storytelling test

The most reliable way to detect unskilled AI use is to ask for stories.

People with real experience can tell stories about it. They can describe a specific moment, a particular failure, a decision they made and what happened next. They can answer follow-up questions because the story is from their actual life. Details emerge naturally because the details are real.

People relying on AI to fake competence cannot do this. The model can summarize concepts, generate frameworks, and produce talking points. It cannot manufacture lived experience. Ask for the story. If the answer drifts back into generalities, you have your answer.

This works in interviews. It works when reviewing proposals. It works in performance conversations. The question — "tell me about a time when..." — has always been the practitioner's filter. AI has made it sharper, not less useful.

Skilled use is its own signature

The opposite is also visible.

Skilled AI use does not produce hollow output. It produces work that is faster, sharper, and more thorough than the operator could have produced alone — but unmistakably the operator's. The AI handled the mechanical parts. The judgment, the structure, the editorial decisions, and the specific knowledge are all the operator's.

Think about a carpenter using an air nailer versus a hammer. A hammer is easy to understand. You know the risks — you might hit your thumb. If somebody showed you a demo of an air nailer, you would be impressed. Walls go up faster. Nails sink cleaner. The demo looks magical.

What the demo does not show is everything you need to know to use the tool safely. The pressure settings. The recoil. The fact that you can damage the material if you misjudge the angle, drive a nail through your own hand, or shoot someone standing nearby. An unskilled operator with an air nailer is not just unproductive. They are dangerous — to themselves, to the work, and to anyone in the room.

AI works the same way. The tool amplifies whatever is already there — competence or its absence. In the hands of a skilled operator, it accelerates good work. In the hands of someone who has not learned how the tool works, it accelerates damage.

The risk that compounds

The deeper concern is not the bad output you can detect today. It is that the operator never develops the skill the AI is performing for them. If someone never learns to write a proposal, structure an argument, or analyze data because the tool does it for them, what happens when they need to exercise judgment the tool cannot provide?

The harder version of this problem is one most organizations are not ready to face: leaders who cannot distinguish skilled from unskilled AI use because they lack the domain expertise to evaluate the work themselves. If you cannot tell whether the output is good, you have a quality control problem AI is making invisible faster than anyone is acknowledging. That is a longer discussion for another piece. For now, the operator-level distinction is enough work.

How to calibrate

Build the habit of asking the second question. Ask the author to walk you through their reasoning, not their output. Ask what they changed from the AI's first draft. Ask what the AI got wrong. Ask for the story behind the strategy.

If they cannot answer, the hand was off the wheel. That is not necessarily a red flag about AI use. It is a signal about competence — and competence is what you were hiring, promoting, or buying in the first place.

The tools make polish cheap. Substance still costs what it always did.