Spain's AI job cuts won't arrive as layoffs — they'll arrive as jobs that never get posted

The number making the rounds is 39.6%: that's the share of Spanish companies that anticipate reducing headcount because of AI, according to the Barómetro de Talento Digital 2026, produced by DIGITALENT Group's Observatorio del Talento Digital and presented by Nacho de Pinedo at the Digital Talent Summit held at ISDI Madrid, as reported by El Confidencial Digital. But the more revealing figure sits underneath it. Among companies expecting a reduction, 35.4% say it will come through less hiring, versus 17.1% pointing to direct cuts, per the analysis published by El Obrero. That is a two-to-one ratio, and it changes the entire shape of the story: the labour market narrows before anything visible happens. No press release, no ERE, no headline. Just vacancies that quietly stop being published.

The exposure is not evenly distributed, and the barometer is blunt about where it concentrates. Administrative staff are flagged as most threatened by 50.7% of companies — five times more than technicians (10.1%) and seven times more than middle managers (6.8%). Senior management registers at 0.5%. Read that last number again: the layer that decides where automation goes sees almost none of the risk landing on itself. The pattern is not uniquely Spanish. The US Bureau of Labor Statistics, in projections published on 27 August 2026 and covered by USA Today, expects office and administrative work to be the category with the steepest percentage decline — roughly -4% through 2035, driven by "the continued integration of automation tools, including AI-based ones" — with cashiers shedding about 200,600 positions, office clerks 156,200 and secretaries and administrative assistants 114,100 between 2025 and 2035. Worth keeping the scale honest, though: -4% spread across a decade is an erosion, not a collapse. It is closer to what happened to bank branch staffing over the 2010s than to a cliff.

The structurally worrying part is what the study itself calls the "junior squeeze". Some 51.7% of companies foresee fewer opportunities for junior profiles and another 31.4% expect their tasks to change significantly; 43.5% identify entry-level roles as the most exposed to AI, against 14.5% who point to older professionals. Here is the trap. The tasks AI absorbs first — drafting reports, cross-checking data, clearing tickets — are precisely the tasks through which people used to learn the job. Automate the apprenticeship and you don't just cut a cost line; you stop manufacturing your own future seniors. De Pinedo, ISDI co-founder and director of the Observatory, frames it in terms of governance rather than adoption, per El Obrero: the question is no longer whether companies use AI, but "how to govern it, how it changes hiring and how professionals learn". A talent pipeline doesn't break at the top. It breaks at the bottom, silently, and the bill arrives five years later.

Then there is the capability gap, which is where the survey turns almost comic. 77.8% of companies report medium, high or very high AI adoption — while 79.2% admit their own professionals are poorly prepared to use it. Only 29.5% have a formalised, communicated framework; 34.8% run on informal guidance or none at all. And 87% of professionals say they would use a personal AI tool if their employer doesn't provide a corporate one. That is shadow AI stated out loud: not a hypothetical compliance risk but a declared behaviour, and the predictable route by which code, client data and IP leak out of organisations. Banning without offering an alternative doesn't reduce usage; it just moves it onto personal laptops where nobody can see it.

Our reading: this barometer is being read as an AI-displacement story when it is really an operating-model story — the same diagnosis we've drawn from the gap between agent adoption and actual profit impact. Companies are cutting the input side of their talent system before building the governance, training and role design that would make the technology pay. One caveat deserves emphasis: these are stated intentions, not measured job losses, and intentions in workforce surveys have a long history of overshooting reality. Still, the direction is consistent with everything we've tracked sector by sector — back-office and routine administration are the most exposed layer; judgment, client relationships and physical presence hold up.

The long view stays intact, and it is genuinely good: work that consists of moving data between systems was never anyone's vocation, and offloading it is how we eventually get to an economy where people do what they're actually good at. But that destination is not reached by deleting junior roles and hoping seniors appear spontaneously. The companies that come out ahead will be the ones that redesign entry-level jobs around supervising AI rather than competing with it, and that hand employees sanctioned tools plus training instead of a prohibition memo. Transition is the hard part. It is also the part that is entirely a management choice.