Table of Contents
A skills gap that isn't closing on its own
Almost half of employers, 47%, say they are experiencing extreme or moderate AI skills shortages, according to Hays' 2026 hiring trends research. That figure alone would be concerning. What makes it worse is the mismatch sitting underneath it: 47% of employers report their people are using AI regularly at work, but only 34% of employees say the same. Somewhere between the boardroom and the desk, the picture of how much AI adoption is actually happening has come apart.
Most organisations have responded the same way: a training portal, a webinar series, perhaps a certification scheme bolted onto existing L&D. Hays found that 50% of leaders now plan to invest in AI and digital skills training. On paper, that should close the gap. In practice, it rarely does, because generic upskilling treats AI capability as a course to complete rather than a way of working to build into roles from day one.
Why generic upskilling underperforms
The employee side of Hays' data is telling. Only 26% of professionals say they have received no support at all from their employer to adopt AI at work, yet 65% say they would take part in AI workshops or training if given the chance. The appetite is there. What's missing isn't enthusiasm, it's structure: training that sits alongside real tasks, with real feedback, rather than a one-off session disconnected from the job itself.
Half of employees also report some level of concern about AI's impact on their own employment. Training programmes that don't address this directly, that don't show people how their role changes rather than disappears, will struggle to land even when the content itself is good.
The entry-level trap
The sharpest version of this problem is showing up at the bottom of the pipeline. Research highlighted by the World Economic Forum this year found that junior-level job postings fell 7% year-on-year in 2025, while senior-level postings rose 4%. Recent graduate unemployment hit a three-year high of 5.7% in the final quarter of 2025. More than a third of young workers, 37% globally, are now in occupations with medium to high AI exposure, and 28% of entry-level workers believe half or fewer of their current skills will still be relevant within three years.
The instinctive response for many firms has been to quietly shrink graduate and junior intakes, on the logic that AI now does the work those roles used to cover. It's an understandable reaction. It's also the wrong one. Entry-level roles are where judgement, institutional knowledge and the instincts a senior hire eventually relies on actually get built. Automate them away entirely and the pipeline for future senior talent erodes with them.
A better model: redesign, don't remove
Indeed's Hannah Calhoon has proposed a practical alternative that's gaining traction: a "human first" rule, where junior staff research and draft manually before bringing AI in to refine; shifting some entry-level work towards a review or "editor" function, where employees verify and improve AI output; and pairing junior staff with senior mentors on AI-assisted projects rather than leaving them to work through tools alone.
The common thread is that AI becomes the tool a junior employee learns to direct, not the replacement for the experience they'd otherwise have gained. That distinction is the difference between a skills gap that closes over eighteen months and one that quietly widens every quarter.
What this means for hiring, not just training
Closing the AI skills gap is as much a hiring decision as a training one. It means vetting candidates, at every level, for AI-native competency alongside technical depth: can they direct a tool, interrogate its output, and know when to override it, not just operate it. It means building entry-level roles deliberately around that collaboration model rather than assuming AI will simply absorb the workload. And it means treating the 65% of employees who say they'd take up training if offered it as a resource to be structured, not a statistic to feel good about.
This is the space Bearcroft works in daily: helping organisations build teams around genuine AI-native capability rather than a credential that claims it. If your hiring process still evaluates candidates the way it did three years ago, that 47% figure isn't going to move on its own.
Get in touch to talk about how we help clients close this exact gap, at every level of seniority.
