Reframing 'talent scarcity': what our Delphi study found

"We can't find people with the right AI skills" is something you hear from almost every employer right now. The usual reading is a supply problem: not enough graduates, not enough training, so produce more of both and the gap closes. It's a tidy story. It also, my co-authors and I came to think, gets the problem slightly wrong.

So we went looking for what experts actually expect. We ran a two-round Delphi study in Finland: you poll a panel, feed the anonymised results back to them, and let them revise. Thirty-six experts from the technology field took part, and the work is out now in Futures. The question underneath it was simple to ask and hard to answer: how should higher education and industry work together to prepare people for an AI-saturated working life, and what happens if they don't?

Why "scarcity" is the wrong word

The phrase talent scarcity makes it sound like talent is a fixed quantity we're running low on, the way you'd run low on a raw material. But the thing employers are short of isn't really a headcount of people who once passed an AI course. It's a workforce that can keep learning as the tools shift under their feet. Call it AI literacy: held by enough people, kept current.

That reframing matters because it changes who is responsible. If the problem is a shortage of graduates, the fix lives inside universities. If the problem is a workforce that has to stay literate in a moving target, the fix lives in the space between universities and employers: in how they share work, learning, and people over time. Our experts kept returning to two levers in particular: how AI gets regulated, and how seriously higher education and industry actually collaborate rather than just talk about it.

Four ways 2040 could go

Rather than make a single prediction, we used the panel's input to build four scenarios for 2040. They aren't forecasts so much as sketches of where things could end up, depending on how those two levers play out.

Talent drought. The bleak case. AI-literate professionals stay in short supply, and the gap between what work needs and what people can do widens instead of closing.

Limited integration. Here the brake is regulation: restrictive rules make it hard to prepare an AI-literate workforce at the pace the technology moves, so good intentions stall.

Stuck in silos. Higher education and industry never really learn to work together. Each does its own thing, the handoffs fail, and people fall through the gap between a degree and a job.

Collaborative intelligence. The one worth aiming for. Supportive regulation and genuine higher-education–industry collaboration combine, and the system actually manages to keep producing people who can work with AI rather than around it.

Laid out like that, the lesson is pretty plain: the good future is the one where the rules help and the institutions cooperate, more than the one with the most graduates or the cleverest technology.

It was never about effort

What stays with me is that none of the three worse scenarios is a story about not trying hard enough at education. They're stories about coordination: regulation that helps or hinders, two institutions that either meet in the middle or don't. The talent is downstream of all that.

I'll be honest about the limits. A Delphi study tells you what an informed panel expects, not what will happen, and our panel was Finnish and drawn from technology, so the view is partial by design. Scenarios are tools for thinking, not promises. But if you take only one thing from the study, let it be this: when an employer says it can't find AI talent, skip "where are the graduates?" and ask what would have to be true, in the rules and in how schools and companies work together, for that talent to still be there down the line.

The full study is open access in Futures.