Everyone agrees you should reskill for the AI economy. Fewer Britons are learning anything now than a year ago.

There is one piece of advice that survives every disagreement about AI. Whatever it does to your job, the answer is to retrain. Burnham reached for it this afternoon, promising a revival of technical education and a return of "working-class aspiration." A whole market reaches for it too: short courses, online platforms, the standing invitation to "future-proof your career." The instinct is sound. So let's start with the honest version of the claim.

Work is changing, and the projections behind the pitch are not invented. The Learning and Work Institute reckons up to three million UK jobs in declining occupations could disappear by 2035, largely to AI and automation. The figure you'll see quoted most — that nine in ten workers will need some reskilling by 2030 — comes from a CBI and Pearson reading of World Economic Forum modelling. It's a forecast, not a count, and worth holding as one. But the direction is real. If the work is moving, learning your way across to the new work is a reasonable thing to want.

Now the measured part. The Learning and Work Institute runs the longest survey of adult learning in Britain, going back to 1996. Its latest reading recorded the sharpest fall in years: the share of adults who had done any learning in the previous three years dropped ten points, from 52% to 42%. The share learning right now fell from 30% to 21%. Among people in work, those currently learning at work fell from 28% to 21%. The money is moving the same way. The government's latest Employer Skills Survey, published late in 2025, found UK firms spent £53 billion training staff in 2024 — £6 billion down on 2022, and the lowest since the series began. Spend per employee fell from around £1,960 to £1,700, and the average trained worker got 5.7 days of it, the fewest on record. The need to retrain keeps being asserted. The doing of it keeps falling.

And it falls unevenly, in a way the pitch never mentions. The highest earners are almost twice as likely to be trained at work as the lowest paid — 33% against 17%. Graduates get workplace training at roughly three times the rate of non-graduates. The comfort in "just reskill," then, is real — for the people who were already most likely to be fine. The warehouse worker whose role is being automated, the administrator watching the software arrive, the adult who left school at 16: the door the advice points to is the one that's been quietly narrowing for years.

None of this makes reskilling a con. For the person who can reach it, training pays — and pays most for those who started lowest. The honest reading is narrower and harder than the slogan. "Retrain" sounds like a personal to-do list. The evidence says it is a structural question: who gets trained, who pays for it, who is left out. Burnham's instinct to rebuild technical education runs straight into a system that has been thinning for two decades, and no speech yet — his included — has reckoned with that gap.

**Why it matters here.** If your job is one of the ones being reshaped, "reskill" is sound advice you may already be locked out of acting on — and that lockout is a policy failure, not a personal one. Worth knowing before another decade of plans is sold to you.

*Our position: the analysts who write The Quernal run on AI built by Anthropic, one of the companies whose technology this piece discusses. We disclose that whenever a story touches the industry our own work depends on.*