Artificial intelligence (AI) may not spell mass job losses for clinicians, according to a perspective published on 12 September 2026 in The New England Journal of Medicine by Dr Dhruv Khullar of Weill Cornell Medicine.
From displacement to demand
Dr Khullar, an associate professor of population health sciences and practising hospitalist, surveys economic theory and historical examples to argue that while AI will change clinical roles and could displace some tasks, it also has the potential to expand the overall clinical workforce in the long run. He draws parallels with medical advances such as cataract surgery and joint replacement, which became more widely used as procedures were made safer, faster and cheaper — an illustration of the Jevons paradox, in which efficiency gains increase overall consumption of a resource.
“But economic theory and history offer distinct reasons to believe that in the long run, adoption of AI agents… could lead to an expansion, rather than a contraction, of the clinical workforce,”
The article addresses widespread fears that AI will replace clinicians in highly exposed specialties such as radiology, primary care, pathology and psychiatry. Dr Khullar accepts that some roles will change or disappear, but warns policymakers and employers against assuming a fixed quantity of clinical work — a mistaken belief described as the “lump of labour” fallacy.
How AI could increase clinical work
Khullar sets out several mechanisms by which AI might broaden clinical activity rather than reduce it. These include lowering the marginal cost of delivering services, improving throughput by making tasks more efficient, and enabling new forms of care that were previously impractical or uneconomic. If AI tools are priced near marginal cost, they could make routine interventions cheaper and thus accessible to more patients.
- Efficiency gains could reduce clinician time per case and thereby increase total capacity;
- Lower cost of care might expand demand, as seen historically with certain surgical procedures;
- New tasks and services enabled by AI could create roles focused on oversight, interpretation and implementation of AI-driven care.
The perspective is careful to acknowledge uncertainty. It does not claim that all specialties will see workforce growth or that harms will not occur. Rather, it suggests that historical patterns of technology adoption in medicine provide reasons to expect that AI could increase the scale of clinical work under certain conditions.
Evidence, limits and implications for policy
The argument rests on economic reasoning and analogy with past medical innovations rather than new empirical data demonstrating that AI has already expanded clinician numbers. That distinction matters: correlation in historical examples does not guarantee the same outcome for a general-purpose technology such as AI. The piece therefore functions as a theoretical counterpoint to more alarmist predictions that assume inevitable job losses.
For policymakers, regulators and health service leaders, the perspective highlights two practical implications. First, workforce planning should allow for role transformation and new types of clinical employment, rather than only modelling reductions. Second, pricing and reimbursement policies for AI-enabled services will influence whether efficiency gains translate into wider access and higher demand.
| Example | Effect observed |
|---|---|
| Cataract surgery | Increased uptake after improvements in safety and efficiency |
| Joint replacement | Expanded provision as procedures became less burdensome |
Dr Khullar’s perspective does not eliminate the need for careful evaluation of AI’s clinical and economic impacts. Peer-reviewed studies, real-world pilots and robust monitoring will be required to determine whether the theoretical mechanisms he outlines actually play out in health systems. Until those data exist, projections about job losses or gains remain contingent.
As health services prepare for digital transformation, the debate prompted by this article underscores the importance of planning that balances risk mitigation with opportunities for expanding access and new workforce models. Whether AI ultimately leads to more clinicians, different clinicians, or fewer clinicians will depend on choices about technology design, regulation, payment models and training.
