The trend: Practicing physicians and AI healthcare enthusiasts sharply disagree on AI’s future role in medicine.

A JAMA article published this month argues that AI already outperforms physicians at diagnosis, prescribing, and chronic disease management. Authors include venture capitalist Vinod Khosla, health tech CEO Neal Khosla, and oncologist Dr. Ezekiel Emanuel. They predict AI will continue to surpass doctors on these critical decision-making tasks as the tech improves and physicians face AI-induced deskilling. They also argue that keeping humans in the loop may ultimately worsen clinical performance, as AI on its own could outperform both physicians and physicians using AI in diagnosing and treating patients.

And if you ask physicians, they’ll say AI’s optimal role is a clinical assistant—not a decision-maker, per a new Sumo survey.

• Not a single physician said AI should independently manage routine clinical decisions without physician oversight.
• 99% say physicians should retain final authority over AI-assisted recommendations.
• 92% believe AI’s greatest value is reducing administrative burden, rather than making clinical decisions.

Why it matters: This perspective is among the first from respected healthcare experts to argue that AI could deliver better medical care than either doctors alone or doctors using AI.

This is unlikely to sit well with working clinicians, most of whom now use AI in their clinical workflows, often daily, for supportive tasks. For instance, 39% of physician AI users use it to pull summaries of medical research or standards of care, 30% use it to create treatment plans, and 28% use it to document patient visits, per a March American Medical Association survey. Just 17% use AI to assist with diagnoses—and even then, it supports rather than makes the final call.

Current AMA guidance keeps physicians central to clinical care, framing AI as augmented intelligence that supports rather than replaces human judgment. Responding to the JAMA piece, AMA CEO John Whyte argued that medicine is more than a collection of tasks: accurate diagnosis and treatment matter, but so do human judgment and intuition.

Implications for healthcare AI companies and providers: As much as they invest in AI, healthcare provider organizations are unlikely to embrace AI-driven clinical care without a human in the loop. There are too many unresolved questions around malpractice, especially since liability frameworks generally assume that physicians remain responsible for exercising independent clinical judgment when using AI.

Still, the rise of clinical AI proponents among respected healthcare leaders could heighten clinicians’ fears that AI-driven deskilling will erode their expertise. Health systems should frame AI as expanding clinicians’ capacity by reducing cognitive and administrative burdens without replacing their most consequential skills. They should also reinforce clinicians’ role in clinical AI, giving them a say in where it’s used, what remains physician-led, and ensuring they can override recommendations.