Overqualified Was Never About Skill
Overqualified almost never meant someone couldn't do the job. It was a word about price, dressed up as a word about skill. Now that the first draft costs almost nothing, the trade behind it has changed.
The page looked finished, and it promised a price nobody had agreed.
It was an offer page – the one page with the number on it. I had given an AI the notes from a client call and let it write the page alone. It came back clean, warm and well laid out. It had also invented a fixed price, a delivery time and a set of terms. Nobody had said any of them.
Nothing on the page looked wrong. That was the problem.
Anyone who has spent years around drawings knows this kind of error. It is the drawing that looks immaculate and carries the wrong tolerance. You only catch it if you have seen what happens when that tolerance reaches the machine.
Overqualified is an odd word when you look at it closely.
It almost never means the person can't do the job. It means the opposite, said politely: you know too much, you will cost too much, you will get bored, you will leave. It is a word about price and fit, dressed up as a word about skill.
For decades the logic held. Most work was the doing – the drafting, the calculating, the first version. You paid for hours of doing, and experience was an expensive way to buy hours. Why pay for twenty years of judgment to produce a first draft that someone three years in could produce?
That trade is what changed.
The first draft now costs almost nothing. The page, the code, the bracket, the report – an agent produces something that looks finished in minutes. What it can't produce is the knowledge of which finished-looking thing is wrong.
The labour data has started to show the shape. Stanford's Digital Economy Lab, working from payroll records for millions of US workers, found that 22 to 25-year-olds in AI-exposed occupations now stand 19 percent below where their employment would be had it kept pace with less-exposed peers. Experienced workers show no comparable gap.
Read the other way, that is a statement about what experience is worth when the doing gets cheap. Entry-level work was mostly the doing. Experienced work was mostly knowing when the doing was wrong.
That knowing used to be overhead. It is becoming the product.
There is an uncomfortable part, and it cuts against the people who would like this story most.
Experience only counts if it is the deciding kind. A lot of what experts spent their years on was doing too – the drawing, the formatting, the reconciling – and that part is now priced exactly like a junior's. Twenty years of doing a thing is not twenty years of judging it.
And judgment doesn't announce itself. A fluent page written by an agent looks just as qualified as one written by someone who has watched a hundred go wrong. If you can't point to the decision you made – the price you refused to let the machine set, the promise you struck out – nobody can see the difference. Sometimes not even you.
Then the honest caveat. Nobody knows how long the gap holds. It is a few years of data in a field that moves monthly.
Factories have an old practice for this. It is called a hold point. The work stops at a defined step and waits until a qualified person signs. Not because the work is probably wrong, but because if it is wrong, someone with the right to say so has to have looked.
After that offer page, I gave the AI one. It may write the page. It may never set the price, and it may never make a promise the owner did not make. Those stay with a person, every time.
That small rule is most of what being experienced means now. The machine does the doing at whatever speed you like. The person stands at the hold point, knows what a wrong tolerance looks like, and signs or doesn't.
The people once told they knew too much are the ones who can stand there.
Overqualified was always a price tag, never a verdict.