Leadership1 distinct publisher3 min readPublished
Graduates routing out of AI-exposed entry-level work are arriving at trades that cannot train them at volume. British bookbinding already shows what a generation without apprentice schemes leaves behind.
The Board Room · Leadership desk
Compiled by The Board RoomSomething wrong?How this is made
Eighty applications per place is the ratio that intake implies [1], and in a hiring process that number tells you the binding constraint sits on the workshop's side, not the candidate's. Ludlow's chief executive states the mechanism plainly: the firm trains its own people because it cannot recruit at volume, there being no courses to recruit from [6]. Interest arriving at a workshop door outstrips the capacity to absorb it, and every apprentice taken on is paid for in a senior craftsperson's working hours.
Bookbinding is useful here precisely because the experiment has already run to completion. Training dried up through the 1950s, 1960s and 1970s, formal apprentice schemes collapsed in the early 1980s, and the industry then got no training at all [3]. That is more than four decades without a formal route in [2]. Paul Kidson's estimate of fewer than 50 handcraft-operating bookbinders left in Britain [c6b] is what sits at the end of that period. It is a small number in a small trade, and it is also the clearest available reading of what a stopped pipeline produces given long enough.
The simple reading is that AI is pushing graduates toward the trades and the trades are pleased to see them. That reading leaves out what labour-supply stories usually leave out. The Guardian's evidence for the migration is a set of interviews plus two employers' application counts [1][2], not a measured flow. We do not know how many graduates have actually switched, what they were earning before, or how many of those 160 applications came from people who would have applied in any year.
160 applications is a rounding error against a graduate cohort, and on volume that is true. The transferable finding is not the size of the flow but the conversion rate. A sector that stopped training for a generation cannot turn enthusiasm into competent workers inside one hiring cycle, however many people knock. The same conversion problem runs in the other direction for accounting, law and software if junior intake thins for long enough [4], and nobody in those functions is closer than five years to finding out.
So name the trade-off as it actually presents. Cutting the graduate rung lowers the cost base in the year you cut it, and the invoice arrives when mid-level roles previously filled by promoted juniors have to be bought on the open market instead. That is a five-to-ten-year exposure rather than a this-quarter one, which is exactly why it survives every budget round.
On the demand side, the money to rebuild capacity is coming from price, not policy. The Folio Society, which says it does not use AI in its art and partnered with Ludlow on a limited edition [5], sits among high-end publishers betting that buyers will pay more for objects made by hand [7]. That funds apprentice places at the pace customers choose to fund them. One of the Cox London apprentices, Nazifa Tabassum, expects AI to change her work rather than replace it, treating it as a tool that assists [8], which is the more durable framing than "AI-proof". For anyone setting a 2027 graduate number, the operative figure remains 80 applications per place: it describes a trade that cannot buy its way back to capacity within a hiring cycle.
Ranked by verification strength, evidence, and original report placement.
The Guardian reports that many young workers have responded to AI disruption by turning to what they see as "AI-proof" heritage crafts such as boat-making, metalwork or bookbinding, hands-on trades requiring specialist knowledge, creative judgment and dexterity.
Cox London, a design house specialising in artistic metalwork, started its apprenticeship scheme last year and this year received more than 160 applications for two apprenticeship places, according to co-founder Nicola Cox.
Paul Kidson, chief executive of Ludlow Bookbinders and a fourth-generation bookbinder, said training in British bookbinding started to dry up in the 1950s, 1960s and 1970s, that all formal apprentice schemes collapsed in the early 1980s, and that the industry then got no training at all.
Joanna Reynolds, chief executive of the Folio Society, said "We do not use AI in terms of our art"; the publisher partnered with Ludlow Bookbinders for a limited edition Game of Thrones release.
Kidson said: "We have to train our own people because you can't recruit at volume because there aren't any [courses]."
Kidson estimates there are now fewer than 50 handcraft-operating bookbinders in Britain today.
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 27, 2026
Follow any of these and your For You feed starts watching them — no settings page required.
invest
CNN's editorial-board trial balloon is settlement currency, not governance1 distinct publisher
product
AEMO now plans for data centres at 13% of the grid, which changes where they can go1 distinct publisher
build
Three services you can delete: queue, cache and search in one Postgres1 distinct publisher
product
Meta owns the models and the data centres, and still pays Microsoft to rent someone else's1 distinct publisher
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
On-record but single-sourced and self-reported
Every load-bearing figure is attributed to a named executive or apprentice at the affected organisation (Nicola Cox, Paul Kidson, Kath Geraghty, Joanna Reynolds) in one Guardian report. That gives good attribution quality but no independent corroboration: the under-50 binder count and the early-1980s apprenticeship collapse are explicitly estimates and recollections, and the broad claim about entry-level white-collar displacement carries no data at all.
Real but small-scale employer-level uptake
There are concrete, countable adoption signals — a metalwork scheme in its second year with 160-plus applicants for two seats, 728 applications for one National Theatre role, Ludlow running in-house training, and one publisher formally disclosing an AI-free art policy. But these are four organisations against a training base described as nonexistent since the early 1980s and a craft workforce estimated at under 50 binders; nothing in the cluster shows sector-scale intake or job creation.
"AI-proof" framing outruns the evidence
The headline and lead frame heritage crafts as an AI-proof escape route for graduates displaced from white-collar work, but the supporting evidence is a handful of apprentices at three employers, an unquantified displacement premise, and application counts that one interviewee attributes partly to AI-generated applications rather than genuine career switching. The article partially self-corrects — an apprentice says AI will change rather than replace her work, and the trades cannot train at volume — which keeps the gap moderate rather than severe.
Interested employers and premium-brand positioning
The quantified claims come from parties that benefit from telling them: craft employers recruiting apprentices and seeking talent (Cox London, Ludlow, National Theatre) and a premium publisher whose product story depends on human hands and on publicly rejecting AI in its art. Scarcity claims — under 50 binders, dying theatre crafts — also support arguments for training funding and premium pricing. The apprentices themselves have no comparable commercial stake, which moderates the score.
Moderate — clear facts, unclear scale
Confidence is reasonable on the specific, dated, named disclosures (Cox London's intake, the National Theatre's 728 applications, Ludlow's in-house training, the Folio Society's AI stance) because they are on the record and internally consistent. It is limited on everything trend-shaped: one publisher, no independent data, and no baseline against which to judge whether craft apprenticeships are absorbing any meaningful share of AI-displaced entry-level workers.