Will AI Replace Medical Specialists?

It’s 1986. A 62-year-old man with diabetes presents to his local emergency room with shortness of breath. An internal medicine physician moonlighting in the ER evaluates and admits him to the general medicine service. There, his primary care doctor rounds each morning before heading back to his private practice. An echocardiogram suggests severe aortic stenosis, so his PCP calls a cardiologist friend for help.

Today, another 62-year-old man with diabetes presents to the emergency department (it’s no longer just a room) with the same symptoms. This time, he sees an emergency medicine physician (now a distinct specialty), who admits him to a hospitalist service (a new specialty). His PCP is not involved. The hospitalist consults a cardiologist, who then asks another cardiologist, this one specializing in structural heart disease, to evaluate the patient’s aortic stenosis.

This is the story of modern American medicine. We’ve become increasingly specialized, focused on narrower and narrower domains.

Around three in four physicians are now specialists. Specialists account for a continually growing share of outpatient visits and most healthcare spending.

But how much further can this go? And will artificial intelligence—which promises to democratize expertise—change the trajectory?

Specialization Solves Some Problems, While Creating Others

Medical knowledge exploded throughout the 20^th century, giving rise to new theories, tests, therapies, and techniques. The iconic country doctor of the 1920s could only store so much in his little black bag and in his head.

As complexity grew, physicians increasingly specialized, dividing the work more efficiently. Between 1933 and 1941, fifteen specialty boards, including ophthalmology, radiology, and pathology, became formally recognized.

Meanwhile, given their historically scarce expertise, specialists have become gatekeepers, shaping medical training, health system priorities, and compensation in ways that often devalue generalists.