What Happens When You Stop Hiring Junior Engineers
One of my kids has recently graduated with a CS degree. He's been applying for entry-level software engineering positions for months now, and the responses - when there are responses - have not been encouraging.
This isn't a new problem. Junior hiring has always been competitive. But the conversations I'm hearing from hiring managers have shifted. It's not "we need someone with more experience" anymore. It's "we don't really need junior engineers - we have AI for the small stuff."
I've been thinking about this a lot, and I think it's one of the most short-sighted decisions the industry is making right now.
The argument for replacing juniors with AI makes sense on paper. AI can write boilerplate code, handle simple bug fixes, scaffold components, and do it faster than a new grad who's still learning your codebase. If you're looking at a spreadsheet and comparing the cost of a junior salary against the cost of an AI subscription, the math looks obvious.
But the math is wrong, because it's only measuring immediate output and not what you lose long-term.
Junior engineers don't just write code - over time, they become the senior engineers your company needs running its systems.
A junior in their first two years learns things that can't be taught in a classroom or by a language model. They learn that changing the date format in Service A breaks the downstream parser in Service B. They learn that the "quick fix" their manager wants will create three months of tech debt. They learn to look at a Jira board and figure out what actually matters this sprint versus what just feels urgent. They learn that the team in building two has a completely different deployment process and that their PR is going to sit for a week if they don't give them a heads up.
None of that comes from writing code. It comes from being in the room, making mistakes, asking questions, and watching how experienced engineers handle the same situations.
AI doesn't learn any of that. It wasn't in the room when the sprint retrospective uncovered why the release went sideways. It doesn't know why the staff engineer pushed back on a microservice split in last month's architecture review. And it doesn't have a working relationship with the team in building two that makes coordination easier next time.
When a company decides they don't need junior engineers because AI handles the simple work, they're making a decision about right now. But they're also making a decision about five years from now. If you don't hire juniors today, you don't have mid-level engineers in three years. You don't have seniors in five years. You don't have staff engineers in eight.
And you can't backfill that gap with AI, because the skills you need at those levels - judgment, context, institutional knowledge, the ability to anticipate what's going to break before you ship it - those skills come from years of experience that started with being a junior engineer who was given the chance to learn on the job.
I understand the pressure. Budgets are real. Headcount is scrutinized. When someone offers you a tool that can do 70% of what a junior does for a fraction of the cost, it's tempting. But the other 30% - the part where a human being grows into someone who can lead a project, mentor the next generation, and make the judgment calls that keep your systems running - that's the reason you hire them in the first place.
My kid will probably find a job eventually. He's persistent and he's good at what he does. But I think about all the new grads who won't, and what that means for the industry in a decade. We're trading our future senior engineers for a cheaper AI bill today, and I don't think most companies have thought through what that's actually going to cost them.