The AI Agent Will See You Now
By
Leonard Zwelling
https://www.nytimes.com/2026/09/14/health/ai-doctors-medicare-fda.html
One of the very obvious potential benefits of AI technology is employing it for the care of the sick. Most everyone is well aware that access to medical care is a challenge. It is very hard to get in to see a doctor these days, especially a primary care doctor.
If you can even get to see a primary care physician, let alone a physician’s assistant, the cost may or may not be covered by your insurance. And, of course, the third component of health care delivery after access and cost is quality and no one really knows at the time they are consuming medical care whether or not it is any good. I say this only because I have been presented with a number of patient histories of late by younger colleagues that demonstrate an absence of good quality health care on the part of physicians who ought to know better.
The potentially beneficial aspect of AI in health care is its ability to take in critical points of the history and the physical exam, even assisting in compiling that history through direct patient interaction and review of past medical records, and then generating both a differential diagnosis and a prioritized list of lab tests and imaging studies to lock in the most likely diagnosis and even suggest the best treatments. I have recently employed the fabulous AI tool Open Evidence and was most impressed with its comprehensive analysis. It was far better than I at generating a differential diagnosis for a patient with a bizarre collection of signs and symptoms. It also proposed a plan to rule in and rule out all alternatives on the list given the story I was told about the undiagnosed patient.
In fact, I was thinking how, if I were still seeing patients, I would be able to spend time talking with the patient, doing a thorough exam, and working with AI to do what I used to have to do by myself using a library.
However, in this article from The New York Times on September 15, there is an implicit suggestion that AI can replace health care professionals completely. I’m not sure I buy that. Like a stethoscope and an EKG, AI models like Open Evidence are great tools, but sick people need human contact to explain what their problems are to them and what the plans are for treatment. This is even more important in the treatment of psychological ailments. I am not sure I’m quite ready to believe that a machine can analyze a patient’s unspoken affect and mood. (I’m not even so sure that doing this over Zoom would work for me with a human on either side of the camera). How is a machine going to handle a serious physical or mental crisis that arises during a patient visit?
As both a physician and as a patient, I still appreciate being touched by my doctors. The bond between the patient and the healer is a sacred one and that bond cannot be reproduced by a machine. This has been true since the first healer cared for the first patient. That being said, at a time in the United States when there are not enough doctors for all who need health care, if a machine could help patients get care when they have no alternative, if the machines can reduce costs, and if, in most cases, quality does not suffer, I am not diametrically opposed to the use of AI in patient care.
However, seeing that I am far more frequently a patient than a caregiver of late, just say for me, I’d still like to see the doctor, the human one. I have a whole host of good ones from surgeons to dermatologists to an interventional anesthesiologist to my concierge internist. In one way or another, each of them adds the human touch to my care. They cannot be replaced by a machine. Healing is still between one human and another.
