AI Isn’t Taking Entry-Level Jobs. Companies Are Cutting Them and Blaming AI

You’re watching AI close the tickets that were supposed to be yours.

You’re seeing the bug queue that trained the senior engineers you admire, the data pulls, the first-pass drafts, and the grunt work everyone hated, and everyone defended, because it was the way in, disappear.

The fear is that the ladder is being pulled up while you’re standing at the bottom of it. I think the fear is real but aimed at the wrong loss. AI is removing two different things, and only one of them was ever worth keeping.

For me, telling them apart starts in a room full of Excel files.

Welcome to Code Like a Girl, the community where women in tech come to be seen, heard, and championed as they walk this path together.

If you are early in tech and watching the ladder change shape beneath you, this is where you find the women who have climbed it and stayed.

My first tech job, third year of university, was at a company that kept shareholder registries for other companies. The department had a ritual. Every reporting period, there was a roomful of grown people who spent two weeks moving numbers from one set of Excel files into another. By hand. Checked twice, because the regulator did not forgive typos.

I wrote a program. It did the two weeks in fifteen minutes.

“What is this,” said my manager.

“The quarter,” I said.

“The quarter takes two weeks.”

“It did.”

Nobody had asked me to write it. The task on my desk said: retype the numbers. Noticing that the retyping itself was the problem wasn’t on any ticket. It never is.

I’d love to tell you I felt like a genius that day. Mostly I was bracing to be scolded for touching something that wasn’t mine.

Hold on to that picture, because we’re going to take it apart. A junior, a boring task, a program, a nervous silence. Inside it is the answer to a question this community has been asking all year.

Dinah wrote here recently about the broken rung: when AI swallows entry-level work, it breaks the ladder to mastery, because the small tasks were where juniors banked their ten thousand hours. There’s one thread in it I haven’t been able to put down since: what exactly were those small tasks giving us?

Because I keep going back to that registry room, and I don’t think the answer is “training.”

Take my fifteen-minute program apart and ask where each piece actually came from.

The knowledge came from university. Three years of writing code for made-up problems. Abstract, impractical, and the only thing in the building that could kill the retyping. Two weeks of data entry taught me plenty about my own endurance. It was never going to teach me how to redesign the process.

The problem came with the desk. Not from the task itself, but from sitting close enough to touch something that cost real money. No homework smells like a room where eight salaries evaporate into cell B47 twice a quarter.

And the permission? Nobody gave me that. I took it myself and braced for the consequences.

Notice what’s missing from that list: the task. What the retyping really did was get me into the room. It was a socially acceptable excuse for a junior to sit next to an expensive problem.

That’s what the bottom rung was all along. Not a classroom. A seat.

If you’re early in your career, this is the distinction that matters. You don’t need a queue of small tasks. You need something real to know, a real problem within arm’s reach, and the nerve to touch it. Two of those you can build yourself. We’ll get to the third.

When a company says “AI is eating our entry-level work,” it can mean two very different things, and almost nobody separates them.

It can mean AI canceled the ticket. The retyping, the boilerplate, the first pass. Good riddance. The value was never in the boilerplate itself.

Or it can mean the company canceled the seat. The junior’s closeness to problems that matter.

Keep the seat and kill the ticket, and a junior can do on day one what people used to earn after months of dues. But empty the room and print “AI handles the entry-level work now” on the justification, and there is your broken rung. Not the automation. The alibi.

So when you hear that sentence at your company, ask which one is happening. It’s the difference between losing your homework and losing your seat in the building.

“Broken rung” meant something before AI touched it, and the two meanings stack in an uncomfortable way.

McKinsey and LeanIn coined the phrase for the first promotion into management, and they publish the numbers every year: for every 100 men promoted into their first manager job, 93 women. Seventy-four if you’re a woman of color. That rung is about being allowed up, into decisions with your name on them.

The rung AI is eating sits below it: execution.

And here’s the small, bitter detail. Execution is exactly where women were welcome.

The International Labour Organization counted: in rich countries, women are nearly three times as likely as men to hold the jobs most exposed to automation. Administration, customer service, coordination, data processing. The tiers we were finally let into are going first.

One rung broken above us. One dissolving below.

And the habits we were trained in don’t transfer evenly either.

The old rung rewarded compliance: take the task, do it well, don’t touch what isn’t yours.
The new game rewards the opposite: look at the task and kill it.

Guess which of those two behaviors girls get drilled in for twenty years before their first job.

If you think that conditioning has expired, here is this year’s number: 21 percent of entry-level women say their manager encourages them to use AI. For the men at the next desk, it’s 33.

Permission is still rationed. Only the tool is new.

Which is why the permission in my story is the part worth staring at. It felt like stealing. It was actually the job. If you are waiting for someone to hand it to you, the numbers say you may wait longer than the man at the next desk. That doesn’t make the risk imaginary. It isn’t priced evenly, and pretending otherwise would be dishonest. It means the silence around you is rationing, not a verdict.

The waiting is not evidence about you.

If the small tasks were never the school, one question is left standing: where does the knowledge come from once no job pays you to be mediocre while you build it?

From practice at fake stakes.

Every expensive skill in nature rehearses in play before it counts for real, and parts of tech already train this way: security teams raise juniors on capture-the-flag exercises, staged attacks where failure costs nothing, and nobody calls it a loss of rigor.

My daughter runs a shop in Minecraft.

“Why is it more expensive today?” I ask.

“There’s only one left,” she says, without looking up.

Consultants bill by the hour for that sentence.

Two lines of fine print, though. The shop teaches her scarcity, tradeoffs, the fact that systems push back. It will never teach her which customer complaint means a client is about to walk. Games teach the shape of problems; the specifics of a field you still have to practice on something that looks like the field, the way a flight simulator looks like a cockpit.

And second: she’s learning economics because an educated parent stands behind the counter, explaining margins. The game is free. The person narrating it is not. My own story folds the same way. Knowledge someone paid for, a seat someone gave me, permission I could afford to take. None of it is handed out evenly.

Practice is safe because it’s fake. And because it’s fake, it can never finish the job. Somewhere there has to be a first time your name sits under a decision that cannot be undone. A release, a diagnosis, a number a client acts on. That, you can only get from an employer.

So the bargain flips. You arrive with the practice already started, on work nobody assigned, and what you’re really asking an employer for is the one thing practice couldn’t give: the right to be answerable, once, for real.

The senior of the next decade isn’t the person who produces the most. It’s the person willing to sign their name under work they don’t fully control.

If you’re the one deciding what AI replaces on your team, the rule is short: cancel tickets, never seats.

A junior alone with AI output is doing office housework. The same junior sitting next to someone who says “plausible, and wrong, and here’s the smell” is doing an apprenticeship. The difference between those two futures is one seat.

Promote the person who kills the pointless task over the person who performs it beautifully. Then watch who that starts promoting. And pay for the practice, the simulations, the sparring hours, on work time, because “figure it out on your own time” is a filter, and we have decades of data on who it filters out.

Sometimes I think about that registry office. Rows of desks, two weeks of retyping, the printer with its opinions. If AI had walked in, the desks would have emptied, and rightly.

What mattered was the seat the ticket gave me: close enough to notice the real problem, and just barely close enough to touch it.

If you’re early in your career, look for the room where the expensive problem lives. And when you’re ready to put your name under something real, say so before anyone thinks to ask.

If you run the room: cancel the ticket, keep the seat. And don’t wait for your juniors to steal the permission. I had to. It shouldn’t still be the price of a seat.

We're grateful to Valentina for sharing her work here. If you want writing that takes a fear everyone feels and hands you a sharper frame to hold it with, subscribe to her publication Human-Shaped Problems.

Code Like a Girl is the community for women in tech who want to be close to the problems that matter, and close to the people who have sat where they’re sitting.

Subscribe for free and join our Rise Together Thursday chat. It’s a community where women in tech share their latest work, spotlight another woman’s post, and support one another through thoughtful comments and restacks.

You’ll come in with work that deserves more attention and leave feeling seen, supported, and connected.