Can AI Give Every Patient the Doctor They Deserve? w/ Ran Shaul, Co-Founder & CPO, K Health

An old piece of wisdom holds that a doctor should listen to a patient carefully before ever touching them, simply because the patient will tell them the diagnosis. Today's guest built an AI solution around that idea, trained on two million patient records spanning 15 years and 400 million doctor's notes, all to capture the reasoning behind the real diagnoses. Before a doctor ever walks into the room, AI has already asked 25 questions and done the listening in about five minutes. [upbeat music] Welcome to Humanify Health Tech, where we explore how AI is rewriting the rules of healthcare and putting people back at the center where they belong. I'm Brian Chi, senior director of content at Net Health and your host. Today I'm joined by Ran Shaul, co-founder and chief product officer of K Health, an AI-enabled clinical care company. K Health's PatientGPT lives inside health systems like the Mayo Clinic, Cedars-Sinai, and as of last month, Atlantic Health. Ran served five years in the Israeli army before building two companies in advertising and finance. He calls himself a data geek because, in K Health's early days, that's exactly what he was, building the model line by line. Nine years later, the model has powered millions of patient visits. Ran, thanks for joining us today.

Thank you very much, Brian. Really happy to be here. Thank you.

Thanks. So I brought up Sir William Osler there at the beginning because I thought I was struck by how you applied this sort of medical tenet to modern technology. You know, in this case, it's, it's basic wi-wisdom shaped-- That, that basic William has shaped how doctors have been trained for over a century, right? A patient's own account of their symptoms usually contains a diagnosis if a physician actually listens long enough to hear it. But here's my question for you to kick off with. There's another study c-- uh, from Beckman and Frankel in 1984 that found that doctors actually interrupt a patient after, like, 18 seconds, and there's other st- other studies that show that it's even more frequent or sooner than that. So the first question to you is this: Are you actually helping the patient or the doctor here?

You know what? Actually, this is a fascinating question. Thank you for starting with this. We actually believe we help both. So I s- before we started, or while we started K Health, I actually studied the practice of medicine because I did not have a medical background. I am an engineer, industrial engineer, data mining, data geek, as you just said. But I was looking at the diagnostic process, and I realized that no matter how you practice medicine, there is a phase called history taking, which you have, depends on the setting, limited time with the patients to understand the situation. And that area, that part of the diagnosis, is actually done before you even reach a conclusion. And you are trained in medical school not to come biased, not to jump into conclusion, not to go faster than needed, and to get a full understanding of the situation. And that is the thing that we thought when we started K Health. Can we condense that? Can we actually create a way that comes in between the patients and the