Everyone says AI will erase young people from companies. I think the exact opposite: the ones who should be worried aren’t those trying to get in, they’re those already inside.
There’s a story that gets repeated everywhere by now, at conferences, in the papers, in LinkedIn posts: AI will kill junior roles. Companies won’t need to hire young people any more, because the work they used to do is now done by a machine, and so the door into working life closes.
I’m here to argue against that, because I think exactly the opposite.
AI won’t kill the juniors. It will kill the people who don’t deserve their place. It will kill whoever, inside a company, isn’t producing value right now and is living off their position. That’s a different thing, and it’s a distinction that completely changes who should be afraid.
Who AI will actually kill
Let’s make the honest list. Until now, inside companies, one category of people has survived perfectly well while having little to do with merit. People with no appetite, ability or will to improve, who nonetheless carved out positions, sometimes senior ones, for other reasons: contacts, networks built over years, excellent communication skills, an above-average command of language, and above all a long, patient, highly effective run of internal politics.
All real qualities, let’s be clear. But none of them is merit. None of them is value produced.
AI doesn’t forgive. It makes output visible and comparable, and it lines up who produces value and who doesn’t in front of everybody. Whoever produces, produces more with AI and can show it. Whoever has been coasting on position is suddenly exposed: they no longer have the repetitive work to hide behind, and they don’t have the tool to multiply themselves. That’s the profile that gets swept away. Not the hungry twenty-four-year-old: the forty-five-year-old who stopped learning in 2015 and has since managed nothing but their own internal survival.
So why is everyone pointing at the juniors?
Because there’s a data point, and it’s real, and it would be thrown at me immediately. In 2024 and 2025 entry-level hiring in the sectors most exposed to AI (software, consulting, basic research) fell considerably: depending on how you measure, between 25 and 35 per cent, with much higher peaks in tech alone. Let me be fully honest: it isn’t AI’s fault alone, economic caution weighs in too, along with a squeeze on junior hiring that predates ChatGPT. But the direction is that one, and I’m not denying it.
And it gets worse for my argument. The most-cited study of the moment (from Stanford’s Digital Economy Lab, “Canaries in the Coal Mine”, November 2025, built on ADP payroll data) puts its finger right in the wound. In the occupations most exposed to AI, young people aged 22 to 25 have lost between 13 and 16 per cent of employment. Experienced workers, in those exact same occupations, have grown: young developers down 20 per cent and experienced ones up 6, and the same pattern (with different numbers) in customer service and accounting.
Translated brutally: today the data says the opposite of what I say. It isn’t the incumbent shaking, it’s the young. It looks like conclusive proof that I’m flat wrong. (It’s a working paper, and the authors themselves caution that AI doesn’t explain everything on its own, but pretending it doesn’t exist would be dishonest.)
And instead it’s proof that it pays to look more closely, because that figure captures two things at once. The first I’ll grant straight away: the old junior is dying. The photocopy junior, the one who earned their place through raw repetitive work (laying out documents, reconciling, doing the basic review, writing up someone else’s numbers neatly), AI eats that work, and rightly so. That’s a piece of the role going away, not the young people going away. The second (experienced workers growing) looks like it proves me wrong, but it’s a snapshot of a time-limited advantage. Two paragraphs from now I’ll explain why time-limited.
The junior emerging now is a different animal. They’re AI-native, hungry, curious, unafraid of getting things wrong, and they use these tools as a second language. That one, I bet, will have more doors open than before. Yes, it’s a prediction, and I own it entirely: I’m betting against today’s data, because I’m convinced it’s measuring the end of a role and the tail of an advantage, not the end of a generation.
Junior or senior is the wrong question
Here’s the point almost nobody brings into focus. The real axis isn’t young against old. It’s whoever adapts against whoever stopped, merit against rent. And yes, I know that today the numbers seem to say seniority is what protects you: that’s exactly what I’m betting will flip, and in a moment you’ll see what I’m basing the bet on.
Because the people genuinely at risk, over time, aren’t those still trying to get into a company. They’re those already inside who haven’t reinvented themselves, haven’t embedded AI in their own processes, and have accumulated over the years a gap in curiosity and habits that today is hidden behind experience and tomorrow will be exposed.
There’s a precedent worth remembering, because it runs against instinct. When ATMs arrived, in the Seventies and Eighties, everyone wrote off the bank teller. It went differently: in the United States tellers went from around 300 thousand (1970) to around 600 thousand (2010), while nearly 400 thousand cash machines were installed. The mechanism is simple: ATMs lowered the cost of running a branch, banks opened many more of them (up 43 per cent in urban branches between 1988 and 2004), and so tellers per branch fell (from 21 to 13) but the total rose. They were doing a different job, less counting banknotes and more dealing with people.
Intellectual honesty: it wasn’t down to technology alone (banking deregulation weighed in too), and the story then changed again. Since 2010 teller numbers have fallen by around 30 per cent, but what cut them down wasn’t the ATM: it was mobile banking, a far more total automation that removed the need to walk into a branch at all. The lesson holds all the same: automating a task doesn’t automatically erase the person who did it. Often it frees them for the part that counts, if they’re capable of doing it.
That’s exactly the bet I’m making on juniors. AI doesn’t take their job away: it takes away the boring part, and puts them in a position to reach the part that matters sooner.
Why, as an entrepreneur, I’d hire an AI-native junior today
I say this as someone who hires people and pays them. All else being equal, today I’d take a junior with no seniority and outsized AI ability over a senior with twenty years of experience who is stuck in an old management habit and has never let AI into their working day.
And the most convincing evidence I have, I’ll give you about myself. At forty-one I enrolled at university, Economics and Business, and closed the three-year degree in under two. Without attending, fitting it into the gaps between one company and another. Not because I’m a genius (I’m not), but because I had AI as a daily tool, the will to get there, and the head to use it properly. Three years compressed into less than two.
Now flip the question. If a junior student, working in the gaps, can compress a degree like that, why shouldn’t a junior in a company compress the jump to mid, to senior, to super-senior in the same way? In two years, not ten. It’s the identical lever: hunger, brains, and AI embedded in daily life.
The reasons are brutally concrete. The junior costs less. Learns faster. Multiplies across functions: thinking, research, getting projects on the ground, production. And a junior amplified by AI becomes senior in a fraction of the time it used to take, with a productive capacity that in some cases is exponentially higher. It holds in finance, in HR, in marketing, in sales, in production, in procurement, in administration. In every branch of a company, if you put in people with AI embedded in their daily work, that branch changes and improves.
I’ll be honest about the incentive, because I hate arguments in disguise: yes, this is also about cost. An AI-native junior who performs like a senior costs a fraction. But that’s exactly what opens doors rather than closing them. It’s all in favour of whoever wants in and is hungry, and all against whoever is in and has stopped.
Here’s why experienced people are winning today (and why it won’t last)
Let’s go back to the data that proves me wrong, experienced workers growing while the young decline. It isn’t an accident, and it’s the same crack I see in my own reasoning, so I may as well look it in the face. If AI removes the whole boring apprenticeship from the junior, it also removes the training ground: judgement doesn’t come from writing numbers up neatly, it comes from the times you got burned. That’s exactly why the experienced person is worth more today. They have a moat, and it’s made of judgement: they know when the AI’s output is plausible but wrong. The young person, alone with an extremely powerful tool, doesn’t have that suspicion yet. Which is why, in roles where a mistake is costly and irreversible (a term sheet, a valuation, a tax move), you still need someone with the scars acting as a net.
But that moat is exactly the thing AI is eroding fastest. The judgement that used to be built over ten years of mistakes, a hungry AI-native junior accumulates in two or three, because they can simulate, fail and correct at a speed the generation before them couldn’t have dreamt of. The experienced person growing today is protected by a time-limited advantage, not by a permanent rent. And if they stop learning, convinced the scars will do forever, they’re the first candidate to become the free ride I’m talking about. The training ground, in other words, shouldn’t be shut down: it should be rebuilt. And in a few years the safety net will be made precisely of the people who are junior today and learning to judge faster than anyone before them.
So?
The story we’re telling ourselves has the wrong subject. It isn’t “AI against the young”. It’s “AI against the free ride”. And the free ride, inside a company, nearly always has grey hair and a good few years of seniority.
For young people, paradoxically, this is the biggest opportunity in decades. We’re handing them the chance to hugely shorten the time it takes to get genuinely good, and to get good at a higher level than the people before them. The barrier that separated whoever had an idea from whoever knew how to build it is crumbling, and not for those already seated: for those who want in.
I was the perfect candidate to become the grey incumbent I’ve been describing all the way through this article: forty-three, a few companies behind me and the comfort of being able to feel I’d arrived. I chose the opposite: going back to being the junior, the student, the one who still has everything to learn, as I described above. Age isn’t what makes the difference. What makes it is still wanting to learn. AI doesn’t reward the young and punish the old: it rewards whoever rolls up their sleeves and punishes whoever stopped. It always has. It’s just that now you can’t hide it any more.
