My doctor’s office is telling new patients to expect a seven-month wait. That’s not a typo.
Seven.
By the time an appointment opens up, the weather will have turned twice. Clinics near me keep closing, and the ones still open seem short on everything, especially time.
I don’t think this is just a rurual-ish Virginia problem. I think it’s the first chapter of a much bigger story, and AI is about to play a large part in it.
The Numbers Behind the Long Wait
The Association of American Medical Colleges projects that the U.S. will be short between 13,500 and 86,000 physicians by 2036. Part of the reason for that shortage will be age. The same report says 20% of clinical physicians are 65 or older, and another 22% are between 55 and 64. A big share of the workforce is close to retirement, and I can only imagine that the situations they are and will be in are making retirement feel more and more like a viable option.
The doctors still working are stretched. The AMA found that 41.9% of physicians reported at least one symptom of burnout in 2025. That was better than 48.2% in 2023. That’s about four out of ten. In family medicine, the specialty most of us rely on, the figure is 45%.
More patients, fewer doctors. Something has to give.
Your Watch Is Already Doing Part of a Doctor’s Job
Apple’s hypertension notifications received FDA clearance in September 2025. The watch tracks how your blood vessels respond to each heartbeat over 30 days and alerts you if the pattern points to high blood pressure. Apple expects it to flag more than a million people with undiagnosed hypertension in the first year.
It doesn’t take a blood pressure reading, and the FDA treats the alert as a reason for more testing rather than a diagnosis. Apple also says the feature isn’t meant for people under 22, people already diagnosed with hypertension, or pregnant users. Even so, a screening that used to mean booking an appointment now happens on your wrist.
AI Trained Only on Medicine Is Already Being Tested
At EPRI, I’m working on a model trained on energy. I expect someone is building the same kind of thing for medicine.
The early results are impressive, with some caveats. In 2025, Microsoft tested a system called MAI-DxO on 304 difficult cases from the New England Journal of Medicine. It solved 85.5% of them, while the physicians in the study averaged 20%. MAI-DxO isn’t one medical-only model. It coordinates general-purpose ones, including GPT, Claude, and Gemini. The cases were hard teaching puzzles with no healthy patients and no mild complaints, so the researchers said it’s unclear how it would do on routine visits or how often it would raise false alarms.
Utah has already moved past the lab. A state program lets an AI system from a company called Doctronic renew prescriptions for 192 chronic-condition medications. Physicians reviewed the first 250 patients in real time and are reviewing the next 1,000 afterward. In the first five months, the AI recommended renewal in 72% of cases, and reviewing physicians agreed with those recommendations 91% of the time. That data came from Doctronic’s own physicians.
A Stanford legal analysis pointed to thin evidence before the launch and unclear accountability when something goes wrong. An expiring prescription often brings a patient back in for a checkup, so automating refills could remove additional visits.
Insurance Companies Will Eat This Up
Here’s my prediction: I think insurers will push patients toward AI “doctors” because the cost savings will be hard to ignore.
AI is used already on the claims side, and physicians are already uneasy about it. In an AMA survey, 61% of physicians said they worry health plans’ AI is driving up prior authorization denials. The AMA also cited a Senate report that found AI denied claims up to 16 times more often than human reviewers did.
A Medical Kiosk Will Replace the Doctor’s Office
I can picture the machine. A cuff for blood pressure. A stethoscope-like sensor for your heart and lungs. A scanner for your eyes, a blood draw, and a breath sample screened for viruses and bacteria. An AI pulls it all into one profile before you’ve finished your coffee.
A company called Forward tried a version of this a few years ago. It raised hundreds of millions of dollars and put AI-driven CarePods in malls and gyms. In November 2024, it shut down without warning. One industry newsletter reported that it had installed just five of the pods, at about $1 million each to build. Forward also struggled to sign up enough subscribers to cover its costs.
Imagine what it could do now. Imagine what it could do a year from now.
In 2022, the FDA authorized the InspectIR breathalyzer, which detects chemical markers of COVID-19 in under three minutes. In testing, it identified 91.2% of positive samples and 99.3% of negative ones, and the FDA still wanted positive results confirmed with a molecular test.
China Put AI Up Front as Japan Keeps Doctors in Charge
The U.S. isn’t alone in this. Other governments have working versions of these ideas, and they’ve drawn the line in different places.
China has fewer doctors per person than we do, about 1.8 per 1,000 people compared with 2.5 in the U.S., according to WHO figures. A company called “Ping An Good Doctor” responded with unstaffed booths it calls One-minute Clinics. You sit down and an AI doctor asks about your symptoms and history, then offers a preliminary diagnosis. A human clinician on the company’s medical team joins remotely, and a cabinet next to the booth dispenses more than 100 common drugs. The company later reported contracts for nearly 1,000 units across eight provinces. Its founder once talked about building hundreds of thousands within three years. That was a few years ago.
The larger shift happened inside hospitals. After the DeepSeek model became available in early 2025, a medRxiv survey found 261 Chinese hospitals reporting local deployment in just over two months. That was still only 0.7% of hospitals nationwide. Doctors mostly use it for decision support and paperwork, and the officials involved haven’t handed it the diagnosis.
China has also set a boundary. In February 2025, Hunan province barred AI from prescribing medicine, including at internet hospitals. Doctors quoted in that coverage said AI can offer suggestions, but the doctor makes the final call.
Japan took a quieter path. The government committed more than $100 million to build 10 model hospitals where AI handles chart entry and helps read tests and images. The plan leaves final diagnoses to doctors. Adoption has been slow, too. One industry summary says nearly 80% of Japanese medical institutions hadn’t adopted AI medical devices as of March 2023.
Different countries, same shortage, and two very different answers to one question: how much authority does the machine get?
So, where do we go from here?
I expect this “AI/Doctor” shift (at least in America) to arrive in stages.
Refills come first, then screenings, then triage, each one pitched as a convenience. Some of it will be better than what we have now. A system that never gets tired or rushed might catch things an exhausted doctor misses, and a seven-month wait could mean the difference between life and death.
It will also shake up the medical field from top to bottom.
Keep in mind, I come at this as a designer and futurist. One question keeps lingering for me. When a machine gives you a health result, how does it explain its reasoning, admit what it doesn’t know, and send you to a real person when it should? Would different AIs give different opinions and diagnoses? A system can be accurate and still be miserable to use. How it’s designed will shape how patients feel about it, and whether they trust it at all.
