How do people learn? The pathway I took is closing, and companies will pay for it

I did not go to university. I studied part time at a polytechnic, a technical college (Which is now part of University of Hertfordshire, hence my alumni), and trained as an electronics engineer while working as a junior in an electronics company. Then I joined a small company as a networking engineer and learned the rest there, from people who knew more than I did and were willing to let me get things wrong on real systems. Thirty years later I'm a CTO and architect.

I don't tell that story to be nostalgic. I tell it because I keep asking myself one question: where is that pathway today? And I think the answer is worrying, for young people and for the companies that will need their seniors in ten years' time.

The pathway, in three parts

My route had three parts: a paid job at the bottom of the ladder, part-time study alongside it, and a small employer willing to take a chance and teach. All three are under pressure.

1. The junior job. The Work Foundation, polling 1,001 senior UK business leaders via Survation in May 2026, found 36% had cut entry-level jobs for 16-24 year olds in the past year, and 43% said AI or automation investment had reduced such roles. The effect was far stronger in medium and large firms (48% and 46% cut roles) than small ones (24%). I'm relying on press coverage of the poll, not the report itself. Skills England's 2026 annual report adds that graduate adverts fell 45% in 2025 and entry-level adverts 25% (Adzuna data), while being careful to say it's too early to know how much is AI.

2. The part-time route. The House of Commons Library's analysis of HESA data shows UK part-time entrants fell from about 470,000 in 2009/10 to about 235,000 in 2019/20, a drop of 50%. Entrants to "other undergraduate" courses, the sub-degree category, fell by 247,000 (72%) between 2008/09 and 2019/20. The "learn while you earn" door I walked through has been narrowing for fifteen years, well before AI arrived.

3. The employer who teaches. Skills England reports business training spend in England fell from £49.4bn in 2019 to £44.8bn in 2024 prices, around £250 less per employee. The Institute of Student Employers found employers are almost twice as likely to build digital literacy after hiring as to recruit for it, which is encouraging, but only if they keep hiring people to build it in.

Is the apprenticeship the new polytechnic?

Partly. It's the closest modern equivalent: paid work with structured learning. The 2024/25 DfE figures show 353,500 starts in England, up 4%, and digital technology starts up around 17% to about 31,600. Starts at level 4 and above rose 15% to 140,730.

But look at who is getting them. Only 21% of starts were under-19s (about 75,000), and over half (51%) were people aged 25 and over. Level 2 starts fell 7%. Growth is concentrated at higher levels and older ages, which is good for upskilling existing staff and less good for a 17- or 18-year-old looking for a first rung. Apprenticeships are a real route, but they are not yet a like-for-like replacement for the open door I had. Skills England also notes youth apprenticeship starts have fallen 40% over the decade and that smaller firms, 99.9% of the UK total, find the system complex to navigate.

Why not hiring juniors is a long-term mistake

The honest reading of the evidence is mixed. The ISE's 2026 survey of 144 employers found 87% expect AI to reshape graduate and apprentice roles, but only 18% expect more than 10% of roles to be replaced. A DSIT and LinkedIn brief cited alongside the Work Foundation poll found UK hiring overall was down 14% year on year, with entry-level hiring tracking the wider market, and it described the AI link as consistent with an effect but not causal evidence. So don't assume AI is the sole cause. But whatever the cause, the effect on the pipeline is the same, and the risks to firms compound:

• No juniors now means no seniors later. Judgement comes from doing the small jobs and seeing what happens. You cannot hire your way out of this forever: if every firm waits for someone else to train the next generation, the market for experienced people gets thinner and more expensive.
• AI needs supervisors who know what good looks like. The people best placed to review AI output are those who learned the craft the hard way. Remove the apprenticeship stage, and in ten years nobody can tell when the machine is confidently wrong.
• Juniors ask the awkward questions. They expose undocumented processes and unexamined assumptions, which is exactly what an AI rollout needs.
• Small firms are where I learned. The Work Foundation figures show small firms cutting entry-level jobs less. That is a point of pride and an opportunity: small employers can still be the classroom.

What I would do (and what I'd ask you to do)

1. Keep hiring juniors, and change the job. Not "do the routine tasks the AI now does", but "review what the AI produced, check it against reality, and explain what's wrong". That's harder than the old work, and it needs a senior's time.
2. Hire for curiosity and character first. The ISE's employers expect critical thinking, AI literacy, communication and adaptability to grow in importance while routine admin and basic data tasks shrink. Tools change every six months. Character doesn't.
3. Use the part-time and earn-and-learn routes. Pair a junior role with day-release study or an apprenticeship, much as mine was. There are now Level 4 AI and Automation practitioner apprenticeships and ten new AI, digital, construction and engineering apprenticeship units published in April 2026. Ask a local college or training provider to do the navigating for you.
4. Protect the learning time. Two hours a week and a named mentor. It costs less than recruiting a senior from a shrinking pool.
5. Write down what good looks like. AI can only be supervised if someone can state the standard.
6. Build the ramp locally. In Sussex there are colleges, universities and meetups, such as the Brighton Dev.AI community, that will gladly connect with employers willing to offer placements and talks.

A 90-day start for a small team: list the routine tasks in each role that AI could now do; run two of them with AI plus human review; redesign one junior role around review and customer contact; set aside protected learning time.

I owe my career to a company that hired an electronics junior and a small firm that trusted a part-timer with real networks. Someone has to do the same for the next generation. If you're weighing how to build capability in-house rather than just buying tools, that is the other half of what I do: leaving your team stronger when I leave. cheesesharp.com