The Future of Health Care Isn't More Specialists. It's Every Clinician Becoming One.

In 1931, roughly three in four American doctors worked in primary care, nearly all as general practitioners. Today, the American Board of Medical Specialties certifies doctors in nearly 40 specialties and roughly 90 subspecialties.

Over time, medicine specialized because it had to. As knowledge outgrew what any one physician could hold in their head, the profession did the sensible thing and divided the body into pieces: a doctor for the heart, another for the skin, another for the hormones. Then it built a system of referrals, waitlists, and handoffs to move patients between the pieces.

Most of us have lived with the consequences. The woman who sees five doctors for what is really one story. The months-long wait to see the one who actually knows the answer. The towns where that doctor doesn't exist at all.

What Specialist-Level Care Looks Like in Practice

At Midi Health, we set out to break that pattern, and we did. Imagine a patient in her mid-forties, living in a town with no women’s health specialist. She goes online and finds availability for a video visit at Midi within a few days. The clinician she meets on the screen has a specialist's depth behind her: deep, current knowledge of menopause and everything it touches (which, by the way, is every aspect of female midlife health). She listens, asks the questions no software can, and educates the patient as she homes in on individualized solutions. By the end of the day, the woman has answers and a plan, not an appointment slot next year. 

Now imagine the same woman three years later, sitting on a video call with the clinician who has known her through perimenopause. Her blood pressure is creeping up. Her glucose is in the gray zone. Her mother just broke a hip, and she wants to know what that means for her own bones. In the old model, each of those questions is a new referral, a new waitlist, and a new stranger with a blank chart. In the new model we’re building, the guideline-level depth of cardiology, endocrinology, and bone health all sits behind the clinician she already trusts. AI gathers the history, checks the evidence, and flags what matters. The clinician makes the calls, within her scope of practice, and when the answer is a mammogram or a procedure, she knows exactly where to send her and stays with the patient afterward.

That is what’s happening at Midi now, with hundreds of clinicians managing perimenopause, cardiometabolic risk, sleep, mood, skin and bone health as one connected picture, not six separate referrals. We move from treating symptoms into preventative health, and from preventative health into longevity support.

AI is enabling all of it. For the first time, deep, current, specialist-level knowledge doesn't have to live in one head, in one building, in one zip code. It can sit beside any clinician at the moment they make a decision. That means we can finally redraw the map of medicine around the patient, instead of around the limits of human memory. It’s the difference between churning out more specialists and every clinician becoming one: not a new doctor for every question, but the same trusted clinician with more behind her. Researchers writing in Health Affairs Scholar this year painted a promising picture of the “rise of the generalist-specialist,” in which primary care and specialty care stop being two separate tracks and become a spectrum any well-supported clinician can navigate. This is what we’re seeing and enabling at Midi.

The Care Abyss: Why We Can't Train Our Way Out

The need is urgent, because the system we built is also buckling under the demand we’re now putting on it, and the numbers show how badly. What we are facing in this country is not a care gap; it is a care abyss. The Association of American Medical Colleges projects a U.S. physician shortage of up to 86,000 by 2036, including a deficit of as many as 40,000 primary care doctors. Some federal estimates run higher still. Zoom in on the population we serve at Midi—women, primarily between the ages of 35 and 65—and the gap gets even worse: national surveys of OB-GYN residency programs find that only a third offer any dedicated menopause curriculum, and other surveys have found that fewer than 1 in 10 graduating residents feel prepared to manage a menopausal patient, despite the fact that roughly 2 million American women cross that threshold every year.

Physician burnout is among the highest of any profession in the country: A Stanford Medicine-led study found doctors are 82% more likely to experience burnout than U.S. workers in other occupations. The top cited driver? Not the patient care itself, but rather ineffective EHR systems and excessive administrative burden—the exact category of work AI is best suited to take off a clinician's plate.

Now add one more pressure point: Health care information was projected to double every 73 days by 2020. It is not humanly possible to keep up. The medical establishment has recognized this information crisis for over a decade, but until now, lacked the tools to solve it. In a now well-cited paper published in the Transactions of the American Clinical and Climatological Association journal the author wrote: “Knowledge is expanding faster than our ability to assimilate and apply it effectively; and this is as true in education and patient care as it is in research. Clearly, simply adding more material and or time to the curriculum will not be an effective coping strategy—fundamental change has become an imperative.” That paper was written in 2011. Fifteen years later, we have the technology to actually drive this change and, at last, keep pace. 

The AI-driven rise of the generalist-specialist changes the math on the workforce question, too. Nurse practitioners, the fastest-growing occupation in the country, can cover more ground more confidently when guideline-level knowledge that used to require a decade of subspecialty training is built into the tools they use at the point of care, which is exactly what we are building for our NPs at Midi.

AI in healthcare isn’t only about efficiency, though those gains are great, and will only continue to improve (one of our clinicians, after using AI for the first time to complete the documentation of a long and verbose visit, declared it was a life changer and that she “would never work anywhere without an AI documentation system again”). What’s more exciting is the impact AI can have on the care itself. It is the first technology that can synthesize a chart, draft a note, flag a risk buried in a patient's history, and accurately ingest and surface an unlimited amount of information. If a major trial is published on a Thursday, the clinician deciding what's right for the patient in front of her can have that finding, in context, at her fingertips, and choose whether it applies—by Friday. As UCSF's Robert Wachter argues in his book A Giant Leap, published earlier this year, this technology doesn't need to be perfect to be worth deploying. It needs to be better than a system that is already failing people at scale. 

Why the Clinician Stays in Charge

But “deploy AI in healthcare” is not one strategy. It’s a fork in the road, and where you land defines what you build. One path treats the AI as the primary actor: an orchestrator that decides which patient gets contacted, by which automated agent, and when, with a clinician positioned as an escalation path for when something goes wrong. The other path—the one we’ve chosen at Midi—treats AI as something closer to the best-informed colleague a clinician has ever had: it reads faster, remembers more, and never gets tired of checking a guideline, but it doesn’t get the final word. A human being still meets with the patient, listens to the patient, and signs off on every action that reaches a patient. Not because we don’t trust the technology to be accurate—increasingly, it is—but because “most of the time” isn't the right bar in medicine, and because patients don’t just need an answer, they need a person who’s accountable for it, and even more fundamentally, someone to talk to. Healthcare requires an incremental-trust model, expanding what technology is allowed to do only as its track record earns it.

The AI systems we build in healthcare need to do the things software is better at: surfacing a patient's full history in one place, checking a treatment plan against current guidelines, drafting documentation, catching a contraindication a clinician might miss on a busy day. Keeping the clinician in charge doesn’t limit what AI can do for health care; rather, it’s how we’ll continue to earn patients’ trust across all of digital health. 

The stakes are especially high for the population we serve. Women in midlife have spent decades being told their symptoms are “just aging,” stuck on waiting lists that run a year for the few brick-and-mortar clinicians who specialize in their care, or told to see five different specialists for what is fundamentally one connected set of changes in the body. They aren’t asking for more technology or a machine to replace clinicians. What they've been waiting for is a system that finally has the capacity to see them and help them. AI, built the right way, is how we get there—not by replacing the clinician, but by finally giving her enough time, information, and support to do what only she can do. Medicine broke itself into pieces because that was the only way a human mind could keep up. It doesn't have to stay that way. The chance in front of us is to put the pieces back together around the patient, with a clinician at the center.