"Why do we even hire juniors anymore? AI does the grunt work."
I hear it more and more. And I get the logic. Five senior engineers with a good AI tool today ship what ten people shipped two years ago. The boilerplate, the simple integrations, the routine refactoring — the things every junior used to cut their teeth on — now happen in minutes. Junior job postings have collapsed over the past few years. The numbers say it plainly, and the mood in the industry says it even louder.
So why do I still hire juniors?
Because I've seen what comes after that equation. And it isn't pretty.
Apprenticeship wasn't a slowdown. It was a senior-engineer factory.
There's a piece of math that looks brilliant on a quarterly basis and catastrophic on a five-year one.
For years the path was simple: intern, junior, mid, senior. The junior wrote the tedious stuff, and in writing it, came to understand how the code actually works. They made their mistakes in cheap places so they wouldn't make them later in expensive ones. That wasn't a necessary evil of the process. That was the process. It's how you produced the senior engineer who, ten years on, knows exactly why the architecture looks the way it does.
Now we're stripping out the grunt work. The gain is real and immediate. The problem is that nobody is planning for what comes next. When you remove the training ground, you don't create efficiency. You create a talent crisis one level up — except it hits quietly, a few years later, when today's seniors start leaving and there's no one to replace them.
When you remove the training ground, you don't create efficiency. You create a talent crisis one level up.
The head of the world's largest cloud company called the idea of replacing juniors with AI one of the dumbest he'd ever heard. I'm inclined to agree — not out of sentiment, but out of arithmetic.
The junior job didn't disappear. It changed.
This doesn't mean we hire juniors to do the same thing as before. Quite the opposite.
We're no longer looking for someone to write boilerplate — the machine writes that. We're looking for someone who, from day one, learns to steer the AI, to read critically what the AI has written, to catch where it quietly got things wrong, and to explain why. The bar went up. The 2026 junior isn't a slower senior. They're someone learning to think at the system level while the tools handle the typing.
And yes, that's harder to teach. Because it's harder to teach someone judgment than syntax. But judgment is precisely the one thing that stays valuable.
Why we do it despite the math
We invest in juniors because someone has to become the senior who knows when the AI is lying. That person doesn't drop out of the sky at 35 with a ready-made sense for when something smells wrong. They're produced — slowly, through mentorship, with the right to make mistakes, with real work alongside real people.
Companies that stop hiring juniors will look more efficient for three years. Then they'll wonder why they have no one to grow into tech leads, and why code quality is quietly collapsing. We'd rather pay the price now, while it's manageable and while there's still someone to apprentice under.
The hive doesn't survive on the bees that already know the way. It survives because it's constantly teaching new ones.
If you're early in your career and you suspect the market is rougher than you were promised, you're right. But we still believe in apprenticeship, because it's how we produce the people we want beside us ten years from now. [See our junior roles] and come in while there's still someone to learn from.