I write this text with unease, because it contains an uncomfortable calculation. So, first things first: this is not a recommendation to replace anyone. It’s an observation — and a worry.
Let’s start with the calculation, because behind closed doors it’s being made anyway.
The uncomfortable calculation
In my work I experience AI at roughly the level of a junior developer — at least on reasonably well-bounded tasks. For that it costs a fraction. A subscription runs to a few hundred euros a month. A junior costs — gross, plus the employer’s social contributions you must never forget in Germany — a multiple of that. And while the human works from eight to five, with a lunch break, the machine runs around the clock. Only, its result needs someone to check it — and a nudge when it can’t get any further on its own.
Not in depth, but in volume: in pure output it delivers what would otherwise take several juniors. I don’t say that with enthusiasm. For me the question was a different one anyway — not how many, but whether even one is affordable: do I hire someone, or invest the money in AI? For me as a one-person company, hiring was simply impossible. Out of my own work and income there’s no money for a second person who only earns their keep on projects after some time. The upfront investment is the problem. Many firms solve it with loans or investors, but without at least a clear prospect that another person can earn their keep in the foreseeable future, it’s pure gambling.
The old dilemma, only sharper
Training juniors was never a good deal, at least not in the short term. You take someone who (still) can’t do much, teach them a great deal over years, pay them for it — and the moment they’re really good, it gets expensive. Either they come and want double, rightly so. Or they quietly move to a competitor who hires them as a senior without knowing where they came from — to that competitor their new salary is just a number, not a doubling.
The firm that did the training rarely benefits. The question “Why should I even train juniors anymore?” gets asked in many organizations. AI only makes it more pressing, because now there’s a cheap alternative for the junior work.
You could stop here and draw a cool conclusion: no more juniors, then. That would be exactly the mistake.
Why the senior becomes indispensable
Because AI delivers junior work — but it doesn’t judge it. Only someone who has learned the craft from the ground up can do that, someone who has been through every mistake themselves. I’ve experienced it: an error the AI simply couldn’t find for days. Countless explanations, countless attempts — in the end I had to find and fix it myself, because the machine plainly didn’t understand what was wrong.
The senior is the only one who can handle AI’s results with confidence. And precisely for that reason, seniors become scarce and precious. I expect firms to poach seniors from one another over the coming years — a competition for the people who bring the judgment the machine (so far) doesn’t have.
The trap
Do you see the pattern? We need seniors more urgently than ever. At the same time, the very stage that produces seniors threatens to disappear. No one is born a senior. Every senior was once a junior — new to the matter, learning, making mistakes. When the junior work goes to the machine and the junior positions vanish, the source that supplies the scarce raw material dries up.
This is no distant theory. It’s a drift that has already begun — the single, commercially sensible decision that, in sum, destroys something that no one, acting alone, wanted to destroy.
What helps — the direction, not the recipe
I don’t have a finished remedy, and I’d be wary of anyone who promises one. But the direction seems clear to me.
Whoever trains people has to remove the dilemma up front. Tell the junior right at the start: “I’m your manager, not your enemy. If you get good, you’ll earn accordingly here too — there’s no debate about that.” Don’t wait until they want to leave. And as an industry, perhaps even as a society, don’t sacrifice training to the short-term calculation just because there’s a cheap machine right now that pretends to be a junior.
How to do that in detail — as a company, as a market, as an education system — is an open question, and a bigger one than I can answer here. But the alternative is foreseeable: a bitter fight over a shrinking number of seniors, while no one grows up behind them. That should worry us. It worries me.
Disclaimer: This article reflects my personal opinion, based partly on observation and partly on research. I welcome corrections where I have demonstrably got something wrong. I have no time for scaremongering; my aim is an analytical, philosophical look at my own working environment.