My prediction is that AI will lead to the loss of hundreds of thousands of healthcare jobs, but few of these will be among physicians, allied health professionals (AHPs), and nurses, at least in the next decade.

[...]

Wander into the billing department of most large health systems, and you’ll witness hundreds of people optimizing documentation, drafting prior auth requests, and otherwise pushing reams of paper, often into fax machines. Amble over to the quality department, where you’ll see scores of workers reviewing charts for data to populate quality measurement spreadsheets, pay-for-performance schemes, and national registries in areas like interventional cardiology and transplantation. Head over to the population health department, whose personnel are struggling to remind patients that they’re overdue for a mammogram or a colonoscopy. The call center… well, you get the picture.

[...]

While administrative and operational jobs have expanded most rapidly, they’re also the most susceptible to AI replacement. The cuts have already begun. At UCSF Health, a decade ago, we started hiring human scribes to assist our busiest ambulatory clinicians in documenting their patient visits. At the program’s peak, about 100 of our busiest clinicians had a scribe – often a pre-medical student on a gap year. Today at UCSF, thousands of our doctors are now using scribes – but it’s no longer a 22-year-old pre-med supplied by a scribe staffing company; it’s an AI scribe supplied by a company named Abridge. Our clinicians are quite happy with their AI scribes (though some tell me they miss the opportunity to mentor the human scribes), and our expenses are a fraction of what they once were.

[...]

In contrast to the major cuts in administrative staff, I predict that very few doctors, nurses, or AHPs will lose their jobs, at least over the next decade. For nurses in the hospital, much of their work involves administering medications, assisting patients with toileting and mobility, inserting catheters and IVs, conducting complex assessments, counseling patients and their families, and responding to emergencies. For the foreseeable future, it’s more likely that AI and robots will assist nurses in performing their jobs more effectively – enabling them to practice closer to the top of their license – than replace those jobs entirely.

For physicians, even fields seen as highly vulnerable to job replacement by AI (particularly radiology and pathology) have proven exceptionally resilient. If you’d asked me 15 years ago which would come first: our radiologists would be unemployed, or I’d sit in the back seat of a driverless car without giving it a second thought, I would have assumed that the radiologists would be toast. Yet in San Francisco, we can’t hire radiologists fast enough, while Waymos are literally everywhere. (In my book, I describe why radiologists have been surprisingly shielded from job replacement.)