No Junior Devs, No Senior Devs Later: The Slow-Motion Problem With AI and Entry Level Work

A Stanford study did the rounds this week confirming something a lot of people already suspected: AI is hitting entry-level jobs hardest. Not middle managers, not senior architects with twenty years of scar tissue, but the graduates and juniors who used to get their start doing the unglamorous, repetitive stuff that taught them the job. I’ve been in IT long enough to remember being that person, and long enough now to be the one deciding whether a junior gets hired at all. Reading the discussion under that article, one comment stuck with me more than the study itself: if you stop hiring juniors now, you don’t have seniors in fifteen years. You have a gap. And gaps in a workforce don’t announce themselves until it’s too late to fix them quickly.

It’s a strange kind of problem because nobody is lying about what’s happening, not really. Executives aren’t hiding the fact that AI lets them do more with fewer people. Investors like that story. Shareholders like that story. The trouble is it’s a story with a short memory. Someone in the thread put it plainly: the psychopaths at the top aren’t going to screw themselves out of a job, so the senior layer survives while the junior layer gets offshored, automated, or simply never hired in the first place. I don’t think most executives are cartoonishly evil about it. I think they’re doing quarterly maths and the fifteen-year problem isn’t their problem, it’s the next CEO’s problem, or the industry’s problem, or nobody’s problem until it’s everybody’s.

What got me was a comment from someone who’s spent over a decade in enterprise automation, essentially making the point that this isn’t really about AI replacing 100% of a job, because that’s never how automation works. It’s about the amount of human input needed to get an outcome shrinking, bit by bit, until the entry point to a career disappears. Nobody automates a whole job. They just automate the part where you used to learn how to do the job. That’s the bit that scares me, and I say that as someone who is genuinely fascinated by what large language models can do. I’ve watched an AI write a passable first draft of a script that would have taken a junior half a day. I’ve also watched it confidently produce something wrong in a way that would have taken that same junior half a day to notice, except now there’s no junior in the loop to notice it, just a senior person too flat out to check.

I think about my own start in this industry a fair bit lately. I got my first IT job doing the sort of tedious ticket triage nobody wanted, and it was genuinely boring, and it was also how I learned to read a system, to notice patterns, to eventually become someone worth paying more for. If that rung on the ladder gets sawn off, I don’t know what replaces it. Maybe nothing does, for a while, and companies just run hotter on fewer senior people until those people burn out or retire, and then discover the shortage all at once, the way we occasionally discover we don’t have enough nurses or teachers, a few years after everyone stopped training them.

There’s a tension here I can’t resolve neatly, so I won’t pretend to. AI genuinely does remove drudgery, and some of that drudgery deserved to go. I don’t miss manually reconciling logs at 2am. But removing the boring bits of a job often removes the exact bits where learning used to happen, and I don’t think that’s a problem AI itself is obligated to solve. It’s a problem for the people running these companies, and for governments who’ll eventually have to deal with a generation locked out of the first rung. Whether either group is thinking about it with any real seriousness, I honestly don’t know.

I don’t have a tidy ending for this one. I’d like to say the market sorts it out, that scarcity of experienced people in a decade will force companies to invest in juniors again. Maybe it will. Markets have a way of correcting things too late and too expensively rather than not at all. In the meantime I’d rather see workplaces treat AI as something that clears space for junior staff to learn faster, not as a replacement for having them in the room at all. That’s a choice, not an inevitability, and it’s worth saying so before we all shrug and agree it was always going to happen this way.