Last week in the UK, the A-level results landed and just as happens every year, hundreds of thousands of students and their parents/carers began the scramble to pass through one of the final gateways of their education before embarking upon their chosen career.

Though plenty of young people weren’t in that queue at all. Official figures published in May put the number of 16-to-24-year-olds in the UK not in education, employment or training at 1,012,000 - 13.5% of them, or nearly one in seven. The first time it has passed a million since 2013. Hold that number too.

When it was my parents’ turn, it was a world (more or less) of the job for life, and a career was something that stayed consistent, predictable for your working life. You got the grades, you got the job and you probably stayed with the same employer but almost definitely the same industry.

When it was my turn, that had shifted slightly. Whilst I may not have known exactly what job I’d do or where I’d do it, whether there would be work at all was never in question. Thankfully, here I still am, in the same industry and with the benefit of seeing it from lots of different perspectives and experiences.

Two years ago, it was our son’s turn, he got the grades (way better than mine) and off he went to study Computer Science, in what felt not only the most obvious choice (the inevitable outcome of growing up in Nerd Central) but also the smartest choice – my entire life’s experience (and that of my recent ancestors) all cried out – if you understand how the machines work, you will always be useful. Which was, I should say plainly, an advantage I suspect most of his year group didn’t have. Not the grades - the advice. A parent who could tell him what to study, with three decades in the industry behind it.

John’s now halfway through his degree and all of a sudden, that sure bet really is starting to feel like a terrible mistake. Not because anything has happened to him, he has all the attributes that you would think the modern workplace would want, aptitude, attitude and a growth mindset that I wish I’d known about when I was his age. But he looks around, he sees the news about what’s happening to people like him – people with experience and technical skills being cast aside by the tech industry and understandably he worries about his future – “how on earth will I find a job, when those with more experience and knowledge are struggling?”

He’s right to worry. You can see him and his peers looking ahead and asking the same basic question - am I going to be OK? And when young people who learned the skills we all insisted were essential are now being told that might no longer be true, and when a million more are outside education, employment and training altogether, you can certainly see why.

But the story they’re all reading is the wrong diagnosis of a real problem. And the wrong diagnosis leads directly to the wrong response.

Let’s start with the aggregate picture. Yale’s Budget Lab has tracked the American labour market continuously since ChatGPT launched and finds stability rather than disruption: the occupational mix shifting only slightly faster than in previous technological transitions, and that shift beginning before generative AI arrived. The Centre for British Progress found no clear evidence of large-scale displacement in the UK either. Forrester’s forecast is that AI will automate around 6% of global jobs by 2030. Six. Not sixty.

The most revealing number comes from Canada, where Statistics Canada asked businesses what they actually did after adopting AI. Among larger firms that had adopted AI, 68% trained the employees they already had and 52% trained their existing executives. Only 33% hired anyone new with AI skills. Where businesses changed their training or staffing at all, the most common change by far was to teach the people already in the building.

Which raises an awkward question about the news that John and his cohort are reading. Nearly 6,000 senior executives across four countries were asked what AI had actually done inside their own companies over three years. More than four in five reported no effect on employment at all. UK firms put the figure at around 0.14%. Sam Altman’s assessment is that almost every company announcing layoffs now blames AI, whether or not it really is about AI. Cisco’s share price rose 13% the week it cut 4,000 jobs.

So the losses are real. The attribution frequently isn’t. The market rewards you handsomely for saying the machine made the decision, which is a very convenient thing to be able to say about a decision you made yourself.

So look at what firms do rather than what they announce, and a different pattern appears - the opposite of the one John is being shown.

In a working paper, researchers tracked 62 million US workers across 285,000 firms, following employment by seniority inside individual companies. From early 2023, firms adopting generative AI saw junior employment fall sharply against comparable non-adopters - while senior employment carried on rising. The fall was concentrated in the most exposed occupations, and it came from slower hiring rather than from anyone being let go. Stanford’s team found the same asymmetry in different data: early-career declines in exposed occupations, with older workers in those same occupations holding steady or growing.

Economists have a name for this now - seniority-biased technological change. A technology that substitutes for entry-level execution while complementing expert judgement.

But this isn’t something that sits still. Seniority-biased technological change describes where the substitution sits today. As the tools improve, that boundary keeps climbing - which means the question isn’t whether your entry-level roles survive, but how far up the ladder the removals reach before anyone notices they’re a strategy rather than a saving.

Which means the experienced people John watched being cast aside were real losses - but not, on the evidence, to AI. So far they are the group the data shows holding up best. John and his cohort are standing where the pressure actually is. And because it arrives as jobs that were never advertised rather than jobs that were lost, there is no announcement, no headline, and nothing for them to point at.

Baroness Martha Lane-Fox, chairing the London Mayor’s AI and Jobs Taskforce, described the risk as not a sudden shock but a quieter drift - fewer entry-level roles, weaker career ladders, growing inequality. Her Taskforce found London employers describing precisely that: recruitment slowed or frozen rather than redundancies made. In every sector they examined, the step from junior work into professional work was disappearing. Worst of all in financial and legal services.

And their own evidence is honest about the gap. The employer survey underpinning the Taskforce’s work found entry-level hiring broadly stable, with no clear statistical link to AI adoption. Which sounds like a rebuttal until you notice it’s the same finding: the thing described consistently in every roundtable simply doesn’t register in the survey data. That is what an invisible contraction looks like from the inside.

The much-heralded “jobs apocalypse”, it seems, is not so much a bang as a whimper (or a Wimpy, IYKYK).

Regular readers will know the question I like to put to professional services firms: your firm’s future depends on senior partners, and remind yourselves – where do the senior partners come from? Of course, they come from the juniors you hire. Stop hiring juniors and you won’t feel anything this year, or maybe even next. But you’ll feel it in 2033, when the talent pipeline is empty and nobody is left who learned the business by doing the unglamorous parts of it. The “on-ramp” isn’t a cost line. It’s the mechanism by which experience gets manufactured - and it only works as a three-way relationship between new talent, powerful tools, and the mentorship from those with the wisdom to know when the output is wrong.

And this stops being a story about graduates the moment you look past professional services. The Taskforce heard the same rung narrowing in transport, where interviewees described the disappearance of career ladders that had been one of the few routes up for workers marginalised everywhere else in the economy. In the creative industries, where freelance and project work is the main way in, they warned the sector risks becoming even less accessible to people from underrepresented backgrounds. Their own conclusion is unambiguous: the decline of entry-level pathways matters not just for employment but for progression and social mobility. There’s a second-order effect to all of this that deserves a piece of its own, and that’s where I’m going next week.

Two findings give me real hope. First, there are firms redeploying people rather than releasing them - taking the capacity AI hands back and pointing it somewhere better, which is the whole opportunity right there. And second, when the Taskforce examined productivity gains on legal work, the quality improvements landed disproportionately on the lower performers, while some of the strongest saw their quality drop.

Think about what that means for John and his peers. These tools don’t threaten them - they are their biggest possible advantage, because they close the gap between them and someone with fifteen years more experience faster than anything in my career ever did. The only thing standing between them and that advantage is being hired in the first place.

Meanwhile, the firms that chose replacement are quietly buying their way back in. Orgvue found 39% of leaders had made people redundant because of AI, and 55% of those now say the decision was wrong. Gartner expects half the companies that cut customer service headcount to rehire for the same work under different job titles by 2027.

So in three years there will be two kinds of organisation. Those with a full pipeline of people who grew up fluent in these tools and were mentored by those who understand the business - and those bidding against each other for the same scarce mid-career hires, because they let their own on-ramp close while watching for an apocalypse that never arrived.

Which is what I keep coming back to with John. The decisions we made two years ago weren’t necessarily wrong. But whether they play out for us or not is now being settled somewhere else entirely - in rooms he will never enter, by people who mostly haven’t noticed they are deciding anything at all.

It turns out my ancestors were almost right. Of course it helps to understand how machines work, but as we’ve seen, machines change and the understanding of them becomes either commoditised or irrelevant. What matters more is the constant that never changes – it’s never about the machines, but is instead what the humans choose to do with them, and how we can help them make the best possible choices for all of us.

And that is a career choice that all of us can make.