Britain could become better at completing today’s work while becoming worse at producing tomorrow’s skilled workers. If artificial intelligence absorbs more of the routine work traditionally given to junior employees, productivity may rise while the first rung of the career ladder becomes harder to reach.
Skills England’s 2026 Annual Skills Report rightly focuses on helping workers use AI and retrain as occupations change. But its chapters on AI and youth employability point to a different problem: how do people acquire experience when software can already perform some of the work through which that experience used to be gained?
We should welcome technology that raises productivity and reject attempts to preserve obsolete work. But a competitive economy also needs routes for new entrants. Britain needs an entry-level jobs strategy built around paid practice, demonstrated capability and progression into skilled work.
The first rung can disappear before the job does
Occupations are rarely automated in one movement. Individual tasks change first. Research, drafting, administration, basic analysis and routine coding can be transferred to software while the wider occupation remains.
Those tasks may be relatively easy to automate, but they have also served as training. A junior employee learns professional standards by doing work that somebody more experienced checks. Over time, they acquire the judgement needed for harder decisions.
Having spent much of my career in recruitment, I have seen how quickly “entry-level” can come to mean “already experienced”. Employers understandably want recruits who can contribute quickly. But somebody still has to provide the first opportunity through which that experience is gained.
Skills England is cautious about the evidence, and rightly so. Its report cites research suggesting that British firms with high exposure to large language models reduced total employment by 4.5 per cent and junior positions by 5.8 per cent. It also records sharp falls in advertised graduate and entry-level vacancies.
Those figures do not establish that AI caused the decline. Skills England also points to weaker economic demand, post-pandemic adjustment and a longer-running deterioration in graduate recruitment. The evidence does not justify declaring a crisis.
It does justify asking whether established routes into skilled work are becoming weaker before the occupations themselves disappear.
Training alone cannot replace a first job
The Government wants millions of adults to receive basic AI training. Helping existing workers use new tools makes sense, particularly if the alternative is that adoption becomes concentrated among a relatively small group of specialists.
But that solves a different problem.
A short course can teach somebody how to use an AI system or check its output. It cannot reproduce responsibility for real work, feedback from an experienced manager or the accumulated judgement that comes from seeing where theory fails in practice.
Skills England’s own work on youth employability recognises the value of meaningful work experience. It also notes that young people can struggle to demonstrate capabilities acquired outside formal employment because recruiters continue to rely heavily on qualifications and previous paid experience.
That problem becomes more important as AI makes polished applications easier to produce. Employers have more reason to look for direct evidence of capability: structured interviews, practical work, occupational assessment and examples of problems solved.
For somebody without an established career history, the route into work therefore needs to provide both learning and evidence. A training certificate may contribute to that, but it cannot substitute for the opportunity to perform real work under supervision.
Use existing institutions to rebuild entry routes
Skills England already has much of the machinery needed to respond. Occupational standards underpin apprenticeships and technical education. Sector Jobs Plans can identify workforce requirements across industries. Local Skills Improvement Plans bring employers, colleges and training providers together around local labour-market needs.
These mechanisms should be used to identify occupations where the work given to beginners is changing faster than the route into the profession.
Employers should help define the capabilities they will still need as AI use grows. Occupational standards and apprenticeships can then reflect the work people are actually expected to perform.
An accountant may increasingly use AI to produce initial analysis, for example, but still needs to understand the underlying data, identify an implausible result and explain a conclusion. The learning route should reflect that change rather than insisting that trainees reproduce work software can already perform.
Local planning matters as well. Large employers can operate graduate schemes and internal academies. Smaller firms often cannot provide every element of a structured programme by themselves.
Skills England has already highlighted a proposed shared work-placement model involving multiple SMEs in Tees Valley. Similar approaches could allow several employers to contribute different parts of an occupational route, with a college or sector body coordinating the programme.
Public support could help with recognised training, coordination and additional supervision. It should not become a permanent subsidy for employer headcount.
Nor can Skills England do all of this alone. It can influence occupational standards, apprenticeships and skills planning. Decisions about employment incentives or wider financial support would require ministers, DWP and the Treasury.
Pay for progression, not participation
There is a reasonable objection to greater intervention. Employers create jobs, not skills agencies, and taxpayers should not be expected to cover training costs that businesses would otherwise meet themselves.
That should constrain the policy.
Skills England should not become a national manpower planner or tell companies how many junior employees to hire. Its role should be to keep publicly supported training aligned with employer demand, help solve coordination problems that individual firms cannot easily fix and make sure that public money creates credible routes into employment.
Success should therefore be measured through labour-market outcomes. How many participants move into paid work? How many remain employed six or twelve months later? Are they progressing in responsibility, competence and earnings? Do employers continue to use the programme because it produces people they want to hire?
Those measures are more useful than enrolments or certificates issued. They also make it easier to stop programmes that attract public money without improving employment prospects.
The wider starting point is not encouraging. Skills England reports youth unemployment at 16.2 per cent in the three months to March 2026, apprenticeship starts among young people down 40 per cent over a decade, and employer spending on training in England falling by 9 per cent in real terms between 2019 and 2024.
AI did not create those weaknesses. But it may make them harder to ignore.
Britain should pursue the productivity gains AI can offer. If technology removes some of the tasks through which people once learned, however, we need better routes into the skilled work that remains. Government should not protect yesterday’s tasks. It should make it easier for people to become tomorrow’s skilled workers.