Entry-Level Roles: Grow or Gut?

Here is the question I’ve been seeing hotly debated, stated plainly. AI is very good at the work we used to hand to our campus hires. Debugging code, reviewing documents, drafting the first version, pulling the data, cleaning the deck. So one camp says: if the machine does the entry-level work, stop hiring at entry level. Another camp says not so fast, that entry-level people become senior people, and if you stop hiring them you are quietly dismantling your own future.

Both camps have data. That is what makes it a real debate and not a talking point.

The case for gutting

The case is stronger than the doubters admit.

Let’s start with the work itself. Entry-level work is codified, checkable, learned-from-a-textbook work: debugging, document review, first drafts, data pulls. That is precisely what AI does best, and it does it in seconds, at a fraction of the cost, without needing to be onboarded, coached, or retained. When a senior person can now direct an agent to do in minutes what used to take an entry-level hire a week, the entry-level hire is not leverage. The entry-level hire is overhead.

Now add the money. Budgets are tight and getting tighter. If a manager can fund exactly one hire, why spend it on someone who needs months of investment before they contribute independently, when the same budget buys a senior who delivers on day one and can supervise a fleet of agents on top of it? One experienced hire, well-equipped with AI, can now carry the output of the small team you used to need. The math favours seniority, and it favours it hard.

The market is already voting. A Harvard working paper covering 66 million workers across more than 280,000 firms found that at companies adopting generative AI, entry-level hiring has fallen roughly 80% per quarter since 2023. The economists call it “seniority-biased technological change”: entry-level demand drops while senior employment at the very same firms keeps growing. Big Tech companies that once filled about a third of engineering roles with new graduates now fill closer to 7%. When that many rational operators move the same direction at once, the believer says, it is not a panic. It is a signal.

And the clincher: speed is the whole game now, and the people who can actually move fast are not the ones still learning the ropes. To move fast and supervise a fleet of agents at the same time, you need judgment, pattern recognition, and the instinct to know when the AI is confidently wrong. That comes from experience, not from a first job. So why spend months growing someone into that competence when the competitive window is measured in quarters? Hire the experienced professional who can do it today, ship faster than the competitor who is busy running a graduate academy, and worry about the pipeline when you have won the market. It is a coherent, disciplined, and genuinely tempting argument. That is exactly why it is dangerous.

The case for growing

Now the other side, which has data too, and is easy to miss under the doom headlines.

Some employers are hiring more entry-level people, not fewer, because AI makes a well-trained entry-level hire dramatically more productive. Brainlabs, a thousand-person media agency, grew its entry-level intake 237% between 2023 and 2026. Their CEO’s point is sharp: “A first-year who knows how to build and run an AI workflow can do work that used to take a team of five. Instead of shrinking the team, it changes what the team gets asked to deliver.” Research from Ramp and Revelio Labs across 21,000 employers found that firms investing heavily in AI grew headcount 10% over the following two years, with entry-level roles growing even faster, at 12%.

But the productivity case is not even the strongest argument for growing. The strongest argument is who these people are. Students coming out of college and the new generation applying for their first roles are the easiest people in your organization to shape. They have not yet learned “this is how we do it here.” They do not carry a decade of habits that AI has quietly made obsolete. They arrive with fresh thinking, precisely the thing every leader claims to want and most orgs slowly train out of people, and they are very often more AI-native than the seniors who would supervise them. You are not hiring cheap hands. You are hiring the most adaptable minds you will ever have access to, at the exact moment adaptability is the scarcest skill.

And notice what actually drove that 80% drop. It did not happen because AI had already replaced those workers. It happened because companies expected it would. Mentions of AI on earnings calls tripled by mid-2023, and firms pulled back on entry-level hiring almost immediately, adjusting for automation they anticipated rather than automation that had actually arrived. A lot of organizations are cutting entry-level hiring based on what they think AI will deliver, not on what it is delivering for them today. That is a bet on a forecast, and forecasts about AI have not exactly been reliable.

There is also a trap hiding in the money argument. When every manager rationally leans senior with their one open role, the org as a whole tips into top bloat: a widening band of senior people at exactly the moment promotions have slowed and there are fewer rungs to move them up. Everyone optimized locally, and collectively we built something heavy at the top and hollow at the bottom.

Where I land

Here is where I stop being neutral. Slashing entry level is short-term thinking, and it is the wrong solution. I do not think it is close.

Start with the longer game the quarterly numbers do not capture. We do not just hire entry-level people to do entry-level work. We hire them so that, five years from now, they are the seniors. The debugging, the ticket triage, the sitting in the review and watching how a principal engineer reacts when something breaks in production, that was never only about the output. It was the apprenticeship. It was how judgment got built, one small managed failure at a time.

AI is absorbing exactly those tasks. And so the question the industry is only starting to ask out loud: where do the senior engineers of 2032 come from, if we stop growing them in 2026? You cannot download seniority. You cannot hire your way out of it either, because if everyone gutted their entry-level tier at once, the senior market is a fixed pie that only gets smaller. And it is not only engineering. The same logic runs through legal, finance, consulting, and yes, HR. The routine early-career work that AI now does was the same work that produced the people we will one day need to judge whether the AI’s output is any good. Cut the apprenticeship and you weaken the very layer that keeps the machine honest.

So this is what I keep coming back to: what we need these people to do may change completely. The fact that we need them does not. You do not solve a tight budget by cutting the one group that is both your most adaptable talent today and your only senior pipeline tomorrow.

But I am not arguing to protect the entry-level role as it was either. That is nostalgia, and the old entry-level job genuinely is being automated. The real work, the harder work, is to redesign what an entry-level job is. Stop hiring entry-level people to do the tasks AI now owns. Start hiring them to direct the AI, to catch what it gets wrong, and to build judgment through work that still teaches it. Brainlabs did not keep the old entry-level role and defend it. They rebuilt it around AI from day one, and hired more people into it.

If you are shrinking, and many of us are, here is a rule worth holding yourself to. Cut proportionally, not from the bottom. The moment all your reductions come from the bottom, you have done it in the place where it is hardest to see and slowest to hurt. A proportional cut keeps the pipeline intact while you get leaner everywhere. A bottom-loaded cut protects this year’s org chart and quietly hollows out the next one.

And one honest caveat, because this whole debate is built on what we know today. By the next campus cycle the world may have turned over completely. It is entirely possible we look up in a year and conclude entry-level hiring should go to zero, or the exact opposite, that every future hire should come in at entry level and grow up AI-native from day one. I am not certain which way it breaks. What I am certain of is that the decision should be made deliberately, revisited every cycle, and never arrived at by default. The danger is not choosing wrong. The danger is drifting into a choice you never actually made.

That is the choice hiding inside this debate. Not grow or gut. Redesign or default. And defaulting, right now, mostly means gutting.