Do apprenticeships still work?

Apprenticeships work as a way of learning. Whether they still work as a bridge is a different question, and the official statistics from four systems answer it uncomfortably. The clearest single number is German. In 2025 around 84,400 young people who wanted a training place did not get one, the highest since 2010, while 54,400 places went unfilled. The most admired apprenticeship system in the world is failing to connect its two ends at the same time, in both directions.

Germany, where the mismatch is the finding

The Federal Institute for Vocational Education and Training reported around 476,000 new dual training contracts in 2025, down 2.1 per cent on 2024 and the second consecutive annual fall. The supply of places fell 4.6 per cent, the second largest drop since 2009. Graded entry.

Read the two sides together, because either alone tells the wrong story. Unplaced applicants rose 19.9 per cent to 84,400, with 15.1 per cent of applicants unsuccessful, the worst since the aftermath of the 2009 crisis. Unfilled places fell to 54,400, which is the lowest unfilled rate since 2020. So the shortage of places is real and so is the surplus, in the same year, in the same country.

BIBB attributes this to matching: regional and occupational mismatch between where young people are and what is on offer. It makes no claim about AI. What it establishes is that a training system can hold both a queue and a vacancy at once, which means volume is not the variable that matters. The connection is.

Australia, and what happens when a subsidy stops

Australian in-training contracts stood at 320,830 at 31 March 2025, down 7.9 per cent year on year, and the fall reached 11.3 per cent by the June quarter. The damage is concentrated: trade apprenticeships fell 3.2 per cent, non-trade fell 17.9 per cent and then 20.2 per cent. Graded entry.

NCVER attributes the non-trade collapse partly to the conclusion of the Boosting Apprenticeship Commencements subsidy, which closed to new entrants in mid-2022. That is a useful control. It shows entry-level training volume responding fast and hard to a funding decision, with no AI involved, and it shows the response falling on the non-trade occupations closest to office work rather than on the trades.

England, where the numbers rose and the age profile did not move

English apprenticeship starts have grown modestly: 337,000 in 2022/23, 340,000 in 2023/24, 353,500 in 2024/25. The age split is the part worth holding. In 2024/25, 51.3 per cent of starts went to people aged 25 or over, 27.5 per cent to 19 to 24 year olds and 21.2 per cent to under-19s, and that distribution has been approximately unchanged since 2018/19. Graded entry.

Half of English apprenticeships are not a bridge from education into work. They are training for people already in it. That is not a criticism of them, and it does matter when apprenticeship growth is offered as the answer to a youth entry problem.

The level 7 cut, and why the obvious reading of it is wrong

From 1 January 2026 the government withdrew apprenticeship levy funding for level 7, master's-equivalent, apprenticeships for anyone aged 22 or over. Funding continues for 16 to 21 year olds, for care leavers and those with an education, health and care plan up to 24, and for everyone who had already started. Graded entry.

It is tempting to file this as government cutting the ladder. Skills England's own evidence makes that hard to sustain. Of 23,870 level 7 starts in 2023/24, 65 per cent went to people aged 25 or over and 2 per cent to under-19s. The levy was largely funding advanced qualifications for staff already established in careers. Whatever else that is, it was not a bridge into work, and redirecting it towards younger entrants is a defensible policy rather than an obvious error.

And level 6 was not cut. The undergraduate degree apprenticeship, which is the one that actually competes with a university place for a school leaver, is unaffected, and its starts have risen from around 6,400 in 2017/18 to around 26,800 in 2024/25. Anyone arguing that degree apprenticeships have been defunded is describing something that did not happen.

Seventy-four

England shortened the apprenticeship instead. The minimum duration fell from twelve months to eight on 1 August 2025, and foundation apprenticeships launched the same day in construction, engineering and manufacturing, digital, and health and social care, with up to 2,000 pounds for the employer.

The National Audit Office checked what happened next. By April 2026, 74 young people had started a construction foundation apprenticeship, against the department's own assumption of 1,000 for 2025-26. Graded entry.

Eight months of data on a new scheme in one sector is not a verdict, and the shortfall is 93 per cent against a planning assumption the department set itself, in the sector the whole package was built around. The lesson available so far is that making the route shorter did not by itself make employers offer it.

The other end of the bridge is thinning too

Graduate recruitment at the UK's hundred leading employers fell 5.1 per cent in 2025, after 14.6 per cent in 2024 and 6.4 per cent in 2023, with a further small decrease forecast for 2026. That is a fall of 24.5 per cent since 2022, the lowest level since 2012, while applications rose 23 per cent in the first half of the season and have roughly doubled since 2023. Graded entry.

The Institute of Student Employers, surveying 155 employers covering 31,000 hires from 1.8 million applications, reports 89 applications per vacancy and a projected 7 per cent drop in student vacancies for 2026. It also reports that 30 per cent of employers increased student hiring, which is the figure that stops this being a story about uniform collapse. Graded entry.

Globally the backdrop is worse rather than better. The ILO puts youth unemployment at 12.4 per cent in 2025, 67 million people, with a NEET rate of 20 per cent covering more than 257 million. Youth unemployment rose in eight of eleven subregions between 2023 and 2025, and in Northern America went from 8.3 to 9.8 per cent. Graded entry.

An EU agency has now named the mechanism

Cedefop, framing its 2027 symposium with the OECD, wrote this in August 2026:

AI takes over baseline tasks which were in many cases performed by apprentices or apprenticeship graduates who get entry-level roles once their programmes are completed. As apprenticeships continue to expand into new fields and occupations, they may also be exposed to a decline in entry-level openings.

Cedefop, 5 August 2026

That is the argument this research has been making, in the words of the European Union's vocational training agency, applied directly to apprenticeships. Graded entry. It is worth being precise about what it is: a call for papers. Cedefop is asking the question rather than reporting an answer, and it says in the same passage that the contraction appears concentrated in white-collar roles, so the effect will be uneven and occupation-specific.

The UK position is more hedged still. The government's own AI and Future of Work Unit found that hiring has been falling faster in occupations more exposed to AI, while stressing that whether AI is responsible for these patterns remains unclear.

Three things people say about this that the evidence does not support

That degree apprenticeships have been defunded. Level 7 for the over-22s was. Level 6 was not, and is growing.

That internships have fallen globally. There is no official statistic measuring internship volume, anywhere. The available numbers come from a job board, a recruiting platform and an employer trade association, and they disagree: Indeed's posting data shows a multi-year contraction, while the US National Association of Colleges and Employers surveyed 284 organisations and found employers expecting to hire 3.9 per cent more interns in 2025-26. This research does not use a figure for internship volume because no defensible one exists, and anyone quoting one should say which commercial platform it came from.

That consultancies have inverted the pyramid. The junior cuts are real and attributable: KPMG's UK graduate intake fell 29 per cent, Deloitte 18 per cent, EY 11 per cent, PwC 6 per cent, and PwC's UK chief said publicly that entry-level intake went from 1,500 to 1,300. But the senior end contracted too. The Big Four collectively made 179 new partners for 2025, a five-year low, down from 276 in 2022. A shrinking base under a shrinking top is compression, not inversion, and no firm publishes the grade-level headcount that would let anyone verify a reverse pyramid. Commentators in the sector describe a diamond rather than a reversal, and that is the more defensible description.

The green shoot, and what it is evidence of

EY's UK and Ireland consulting managing partner Sayeh Ghanbari has argued that heavy remote work is just not the route to success in the world of AI, that consultants must be good at what we're really good at, which is to be human, and that a career in consulting cannot develop those human skills through so much remote work. KPMG's UK head of advisory has said the firm is experimenting with new approaches to in-person training to support soft-skill development. Graded entry.

This matters as a signal about where senior attention has moved. It is not evidence that anything has been fixed. No measurement accompanies it, no firm has published a commitment to restore the training it cut, and the causal claim underneath, that office attendance rebuilds human skills, has not been tested. The counter-case exists too: McKinsey has said it plans 12 per cent more junior hires in 2026 than 2025 in North America, on the argument that the work still requires the same intellect.

What nobody has measured

No study establishes that firms which cut junior intake have suffered a measurable consequence. Executives predict one, which is not the same thing. The Financial Reporting Council's 2026 review reports audit quality continuing to improve overall while flagging risks from extended team models, and that is the closest thing to a check that exists.

Nor has anyone measured whether an eight-month apprenticeship produces what a twelve-month one did. The duration was cut on the reasoning that occupational competence can be reached sooner where that makes sense, and the evidence for that proposition, in either direction, is not published.

And no official body has established that AI is causing any of it. The measured falls have other explanations available: subsidy withdrawal in Australia, matching failure in Germany, and a weak graduate market in the UK that began contracting in 2023. The AI account is plausible, it is now being made by Cedefop, and it remains unproven.

What follows for anyone building a route in

Germany is the instructive case, because it has the volume and still cannot connect the ends. That argues against treating this as a funding problem alone. The English figures point the same way: starts rose, the age profile did not move, and the shortened replacement scheme drew 74 people in its flagship sector.

The design question underneath is the one this research keeps returning to. An apprenticeship is valuable because the apprentice does work that a more experienced person checks, which is productive struggle and retrieval practice arranged as a job. If the tasks that made up that work are the first ones automated, then the scheme can be funded, shortened, subsidised and filled, and still not do what it used to do. That is the missing rungs problem arriving at the institution built specifically to solve it, and what accumulates when it goes unaddressed is capability debt.

Key sources

• BIBB (2025). Angespannte Lage auf dem Ausbildungsmarkt, 10 December 2025. Graded entry.
• NCVER (2025). Apprentices and trainees 2025. Graded entry.
• House of Commons Library (2025). Apprenticeship statistics for England, CBP 06113, 4 December 2025. Graded entry.
• Skills England (2026). Evidence on defunding of level 7 apprenticeships. Graded entry.
• National Audit Office (2026). Increasing construction skills, 13 July 2026. Graded entry.
• Cedefop (2026). Call for abstracts: apprenticeships in the age of AI, 5 August 2026. Graded entry.
• ILO (2026). Global Employment Trends for Youth 2026, 11 August 2026. Graded entry.
• High Fliers Research (2026). The Graduate Market in 2026. Graded entry.
• Institute of Student Employers (2025). Student Recruitment Survey 2025. Graded entry.
• Financial Times (2026). Junior consultants called back to office as AI increases need for human skills, 27 August 2026. Graded entry.

Related SuperSkills research

The mechanism is the missing rungs, and the accumulation is capability debt. On the labour market: will AI replace entry-level jobs and what should I tell my children to study. For the person inside it: should juniors use AI at all. On the learning underneath an apprenticeship: productive struggle, retrieval practice and the missed reps. On what schools are being told: the official guidance on AI in education. Internationally, AI and work in Asia covers the one government that has written this risk into a national framework.

About this research

Rahim Hirji is the author of SuperSkills (Kogan Page, 2026), keynote speaker on AI and human capability, and founder of The SuperSkills Intelligence Company. He has worked in and around education technology for a decade and wrote on early careers and internships in Minterns taking Minternships? (25 August 2019). Every statistic on this page was read at the issuing body's own publication on 30 August 2026. Where a claim could not be verified, including internship volume, it is named rather than estimated.

How this research works  ·  Reviewed quarterly  ·  Found an error? Tell me and it is corrected on the page.