In Plain Sight - 08
In Plain Sight is a weekly blog from Rhys that we will publish every Thursday/Friday. Each edition will be focused on a thorny topic within investing, startups, learning and work and policy.
In Plain Sight is partially a personal attempt to think independently and to avoid leaning too much on AI for answers to our questions. Will we use AI to edit and polish the text? Yes. But more importantly, will we come up with all of the ideas and analysis? Yes.
Here follows Edition 08 - '1 million young people out of work: is the career ladder breaking?'
------
Almost a million young people in the UK are officially NEET: not in education, employment, or training. The latest ONS figure is 957,000, a rate of 12.8% of 16 to 24 year olds. ONS itself labels these "official statistics in development" because of ongoing survey response-rate problems, so the broad picture is more trustworthy than the precise number.
That total splits into two groups. 547,000 are economically inactive: not working and not looking for work, a category long-term sickness weighs heavily on. BBC reporting on the wider NEET population found that nearly a fifth report a mental health condition and quoted youth workers describing teenagers struggling with everyday tasks, not just job applications. That's a separate topic and not one we try to address here, because this piece focuses on the jobs market. There will be a grey area between this group and the other, smaller group we're looking to discuss here.
The other 411,000 are unemployed in the stricter sense: out of work and actively looking.
We're focused on this group - the one shaped by what's happening on the employer side. Though the headline data is British, the pattern shows up across many countries: entry-level hiring pressure appears to be building wherever AI adoption is colliding with a youth labour market, though as this piece gets into, how much of that is AI and how much is something else is genuinely contested. The evidence below is largely from the British case, and we flag where it's drawn from elsewhere.
Ask an employer why they're not hiring junior staff and AI is the easy answer. The British Chambers of Commerce found that 54% of UK SMEs now use AI tools, more than double the 25% using them in 2024, and separately noted that entry-level hiring costs have risen roughly 7% in real terms once National Insurance, minimum wage, and employment rights changes are accounted for. Cheaper to automate, more expensive to hire: a real incentive, whatever is actually driving the decline.
But the UK government's own snapshot of entry-level hiring is considerably more careful than that framing implies. It found UK entry-level hiring down 14% year-on-year as of spring 2026, with the steepest falls concentrated in roles where AI capability has grown fastest, such as software engineering and graphic design. Then it stopped short of the causal claim: "this is not yet causal evidence of AI's impact," the analysis says, pointing instead to a broader cyclical slowdown and a mismatch between the general, analytical skills graduates offer and the specific, production-ready skills employers say they want.
The direct evidence that AI itself, not just a soft economy, is displacing young workers comes from the US, not Britain. Stanford's Digital Economy Lab, using payroll data covering millions of American workers, found that employment among 22 to 25 year olds in AI-exposed occupations now sits 19% below where it would be had it kept pace with less-exposed peers of the same age. That's a striking number, but it's the American case. Nobody has yet published UK data with the same rigour or the same size of effect. The fair summary is that the mechanism is demonstrated in the US and plausible in the UK, but not proven.
A handful of commentators have started calling this pattern the missing rung: institutions have long relied on low-status, low-risk work to train high-skill people. If you automate that work without replacing the training it provided and you get efficiency today but a shortage of talent tomorrow.
The middle may be thinning too
Many people have discussed the entry-level problem but few have clocked another: there's a case that the rung above it is thinning as well, though the evidence is softer.
A Gartner prediction - now widely recycled across HR and trade press - holds that one in five organisations plan to use AI to flatten their management structure by 2026 and that among those that do, more than half of their own middle-management layer could go.
Separately, research summarised by People Managing People, which draws on largely US and global HR surveys rather than European data, found middle-management hiring fell 43% in 2024, roughly three times the decline in entry-level hiring over the same period and that 41% of employees say their employer has cut a layer of management outright.
This is arguably a re-shaping rather than a removal. One rebuttal argues AI mainly removes the "process enforcer" version of middle management, the layer that chiefly relays information and polices process, while making the "translator" version, the manager who absorbs ambiguity and reconciles what different stakeholders actually want, more valuable rather than less. On that view, the risk isn't necessarily that middle management vanishes, it's that companies cut the half of it that they may need most.
Two kinds of "ready" and the market appears only interested in one of them
There are two ways a business can get a person who's ready for a senior role. It can hire someone ready-made, already shaped by years in a job that may no longer exist in the same form. Or it can hire someone ready to be made ready and commit to the years of supervision, mistakes, and correction that shape them into it.
For most of the last few decades, businesses did both, because the entry-level and middle layers of the org chart made the second option cheap. That's getting more expensive, in the UK's case for a mix of AI, cost, and cyclical reasons that aren't yet possible to cleanly separate. The result is fewer entry level roles available to young people.
That's the plain answer to whether junior people will still need training to do the middle-manager job: yes, obviously. Nobody is born knowing how to run an account or handle a hard conversation with a client. What's under threat isn't the need for that training. It's the rungs people used to stand on while the training takes place.
What could go wrong with this argument
Start with a more basic objection to the whole premise: blaming AI lets a company frame a hiring freeze as about efficiency rather than performance and from outside it's genuinely hard to tell a freeze driven by weak performance from one driven by a real productivity gain. The Economist makes a version of this case at the aggregate level: it estimates AI has been a net job creator in the US, putting roughly 1% of professional jobs - about a million positions - into a new "AI jobs" category, against roughly 200,000 layoffs 'officially' attributed to AI since mid-2023. Most of that million isn't like-for-like with the entry-level jobs this piece is about, though: a large share sits in the data-centre build-out, and the Bureau of Labor Statistics expects utilities to be the fastest-growing big US sector through 2035. A construction and infrastructure boom offsetting a white-collar hiring freeze isn't the same as AI being neutral for young professionals - it just means the losses and the gains are landing on different people in different places.
There's also the evidence-quality gap this blog has tried to acknowledge. The UK's own government analysis is a useful check: entry-level hiring is falling in step with a broader economic slowdown and the analysts closest to the domestic data explicitly say they can't yet separate any AI effect from a soft labour market and a graduate skills mismatch. The clearest AI-specific evidence sits in the US, not the UK, and it would be easy to borrow the drama of the American numbers to describe a British trend that may have different causes.
Long-term sickness, which drives the larger, economically-inactive half of the NEET figure, has little to do with AI either. Treating the whole million as a disguised labour-market story would be exactly the kind of overreach we are looking to avoid.
The middle-management story carries the same caveat raised above: if the reshaping argument holds, the danger probably isn't fewer managers overall - it's cutting the kind of manager a business will need most while keeping the kind that was easiest to automate around anyway. This distinction is hard to observe in macro data.
For employers, policymakers, and the young people caught in the middle
If you're an employer cutting junior and management layers at the same time, it's worth asking directly where the next ten years of your leadership team is meant to come from and which kind of manager you're actually cutting. That's a succession question and most workforce reductions don't take this into account.
If you're a policymaker, whether you're running the UK's Youth Guarantee or an equivalent scheme elsewhere, getting young people into training or a placement is necessary but not sufficient if the roles they're training for keep shrinking at the other end, or if nobody can yet say with confidence which roles that will be. Any national apprenticeship or skills system needs to be judged against where the jobs are likely to be in five years, not where they were five years ago, while being honest that this is a forecast.
If you're a young person navigating this: the ladder hasn't disappeared, but it may have fewer rungs and they may be spaced further apart, especially in occupations where routine information-processing was always most of the job. The skills worth chasing are the ones that were always hardest to automate anyway: judgment under ambiguity, the ability to build trust with a client or a team and the willingness to be the person in the room who actually understands the problem, not just the tool that summarises it.
We think there's a real opportunity here, not just a problem, for anyone building tools and programmes that compress the time it takes to develop those judgment-heavy skills, rather than assuming the old multi-year apprenticeship model will keep working with fewer rungs to stand on. If that's what you're building, we'd love to talk.
