AI Workforce Europe

The narrative is everywhere: AI will solve the workforce crisis by doing more with fewer people. Walk into any health ministry, hospital board, or medical school in Europe and you'll hear it. Some fear it. Some are quietly counting on it to finally fix the budget.

I think we're falling into a trap, as all of this is not Black & White, like the image above shows.

This week, Dhruv Khullar in NEJM makes the economic case that AI could actually expand the clinical workforce. Not just during implementation, but structurally, long-term. It matters for Europe especially where we're facing simultaneous shortages, aging populations, and declining medical school enrollment.

The trap is the "Lump of Labor" fallacy. It's the belief that there's a fixed amount of work to be done. It's wrong. The type and amount of work change. Always have. Medicine in 1980 looked nothing like medicine in 2026, and medicine in 2026 will look nothing like 2080. Yet we keep planning as if workforce needs are static.

History says the opposite: the Jevons paradox. In 1865, economist William Stanley Jevons observed something counterintuitive. Steam engines that burned less coal didn't reduce coal consumption. They increased it. Why? Because efficiency made steam engines cheaper and more useful, so people built more of them, used them more, and demanded more coal overall. The same paradox played out with lighting, air travel, and computing.

In healthcare, it's already happening. Cataract surgery and MRI became faster and cheaper. So volumes went up. Radiologists were supposed to disappear. Instead, there are more of them in the US today than a decade ago, with larger workloads. Efficiency created more demand, not less.

Europe's reality: In the Netherlands, France, and Germany, we have unfilled positions across primary care, psychiatry, and rural medicine right now. That's not because efficiency will shrink those needs. It's because AI hasn't yet saturated unmet demand. The care we're not delivering today. Once it does, the Jevons paradox suggests we'll use it to expand access, not contract workforce.

Then comes the O-ring theory. Complex healthcare systems fail at their weakest link. Name it after the Challenger disaster or call it clinical reality: if one step in a chain of dependent steps fails, the whole outcome collapses. As we automate routine tasks (documentation, initial triage, imaging reads), the remaining human steps become more critical, not less.

A radiologist who once spent 60% of their time on straightforward reads can now spend that freed time on complex cases, teaching, managing diagnostic uncertainty, and clinical judgment. Miss that step and the whole diagnostic chain breaks.

→ The work is not fixed. Unmet need in European healthcare is vast. AI absorbs that first. Then creates new possibilities.

→ Roles will change, not disappear. The note gets written by systems. Strategy, nuance, judgment? Still human.

→ Access changes expectations. When care is available 24/7, volumes rise. Someone coordinates, integrates, and makes meaning from all of that.

→ The coordination step becomes more valuable. Managing uncertain patients, negotiating treatment plans, motivating behavior change, leading teams. This is where humans add irreplaceable value in the O-ring system.

My addition: if AI is infrastructure as opposed to a tool, remember what infrastructure does. The printing press didn't shrink writing. It created whole professions we couldn't imagine before.

The caveat: because I'm a techno-realist, this is not automatic. Roles will change, some will shift entirely. But if we plan European workforce strategy around the lump of labor fallacy, around subtraction, we'll engineer the shortage we fear.

So: is your workforce plan for 2035 built on the lump of labor fallacy, or on redesign for an expanding health ecosystem?

---

Source: Khullar D. Artificial Intelligence and the Future of the Clinical Workforce. NEJM Perspective, September 17, 2026. DOI: 10.1056/NEJMp2607831