AI Didn't Make Juniors Redundant. It Made Seniors Non-Negotiable.

The easy read of the last two years is that AI thins out the junior end. The agent does the entry-level work, so why hire the entry-level person. I understand the logic and I think it leads somewhere expensive, because it mistakes what juniors were for.

Juniors were never just cheap output. They were seniors in progress. The boring work they did was how the judgment got built, one small mistake at a time. Remove the boring work and you have not just saved a salary. You have removed the thing that manufactures the senior you will desperately need in five years.

It is the classic mistake of reading a person as a cost line instead of an investment. The junior is expensive relative to their output today and cheap relative to the senior they become. Cut them because an agent matches their output now, and the saving is real and immediate, and the cost arrives years later, quietly, as a gap where your next senior should have been.

Here is the shift that matters. When a machine can produce plausible, confident, senior-looking output on demand, the scarce thing is no longer the ability to produce it. It is the ability to tell whether it is any good. And that ability, the calibrated instinct for when the tidy answer is quietly wrong, is exactly what seniority is.

So the value did not fall at the junior end and hold everywhere else. It concentrated hard at the senior end. A team full of people prompting tools, with nobody who has the earned judgment to catch the tool when it is confidently wrong, is not a lean team. It is an exposed one.

And the exposure is exactly the kind that does not show up in a demo or a good quarter. A team of prompters can move fast and look brilliant, because the tools are genuinely good, right up until they hit the one problem the tools are confidently wrong about. Then you find out whether anyone in the room has the depth to notice, and depth is not something you can prompt for in the moment you need it.

The problem is that this is invisible for a while. A team of juniors and agents can look productive, even fast, right up until the day something subtle goes wrong and nobody in the room has the depth to see it before it ships. The bill for skipping the senior arrives late, which is exactly why it is easy to talk yourself out of paying it now.

And the pipeline problem sits underneath all of it. If the industry stops growing juniors because agents do the starter work, where does the next generation of seniors come from? We are at risk of pulling up the ladder and being surprised, a few years on, that nobody climbed it.

None of this is an argument against using the tools, or even against smaller teams. It is an argument for being clear-eyed about what a small team costs you if it has no depth in it. A senior is not a luxury on an AI-heavy team. They are the thing standing between confident output and a confident mistake.

If I were advising anyone building a team right now, it would be this: keep hiring the juniors, and be far more deliberate than you used to be about actually growing them, because the on-ramp that used to happen by accident, through the boring work, now has to happen on purpose. Seniors do not appear from nowhere. They are made, slowly, and the making just got much easier to skip.

I say all this as someone who loves the tools and reaches for them constantly, not as someone nervous about them. That is rather the point. You can be entirely sold on the technology and still think cutting your juniors to fund it is a mistake, because the two are not actually in tension. The smart move is to use the tools hard and keep growing the people who will one day be the only thing standing between the tools and a confident, expensive error.

So if you are making hiring calls right now, the question I would sit with is not whether AI replaced your juniors. It is whether you still have anyone with the judgment to know when the machine is wrong, and where you think the next one of those comes from.