If you’ve never played a game of make-believe with one of your in-laws, I highly recommend it.
On a Zoom call a few months ago, Ambience Healthcare did a demo for me and my father-in-law, a pediatrician. Together, he and I played a parent and child in a doctor’s appointment—I pretended to be an ornery teenager, he was the concerned dad, and an Ambience employee played the doctor. We mumbled, I whined about middle school and stomach pains, and someone at one point interrupted the visit to ask about lunch orders.
Then, we all watched as Ambience’s medical transcription technology whipped up a detailed summary of the visit. I asked my pediatrician father-in-law what he thought of the outcome. His response: “Wait, can I actually use this in my practice? I want this.”
Ambience, cofounded by Mike Ng and Nikhil Buduma in 2020, is an AI-powered platform geared towards improving documentation processes in medicine. And, in this case, what my father-in-law and I were looking at was the company’s AI medical scribing technology. A medical scribe, for the uninitiated, has one of the most important and least wanted jobs in medicine—real-time note-taking on patient visits. In a world filled with AI solutions in search of a problem, Ambience is focusing on a pain point that just about any doctor will attest to (after all, who likes filling out paperwork?).
Ambience’s technology has now been rolled out at John Muir Health, a prominent Northern California healthcare system, with 16 specialties accounted for, Fortune can exclusively report. John Muir is sizable, with a network of north of 1,000 physicians across two facilities. Dr. Priti Patel, John Muir Health Chief Medical Information Officer says that her doctors aren’t just adopting the technology, they’re latching onto it. In metrics provided by John Muir, of the system’s doctors 85% say they would prefer to continue using an AI scribe.
“We've been doing all kinds of surveys, and everyone is basically saying that they do not want this to go away,” said Patel. “I think we’re going to get to the place where this is the standard of care. There's really no going back at this point. Imagine getting rid of the smartphone. That’s not happening.”
Ambience is part of a growing wave of venture-backed AI startups focusing on medicine and transcription that are gaining ground. Ambience has raised $100 million to date from backers like Kleiner Perkins, Andreessen Horowitz, and the OpenAI Startup Fund. Ambience’s competitors are also well-funded, especially Abridge, which was founded two years earlier and whose backers include Union Square Ventures and Bessemer Venture Partners. (Abridge also just rolled out across Kaiser Permanente.) Others in the space include Cathay Innovation-backed Nabla and Venrock and First Round-backed Suki. It’s almost like everyone in Silicon Valley has a horse in this race.
Still, I’d be lying if I said I didn’t have some caution about the technology itself—this strikes me as the kind of thing that, when it works really well, it’s great, and when it stops working or starts misfiring, it could be catastrophic. Ng says that the goal for Ambience is to be embedded into clinicians’ workflows, without overhauling existing systems and while implementing “the right safety guardrails, and the right sort of governance.”
Ambience’s origins trace back to MIT, where Ng and Buduma first met. They later became friends with the early teams at Google Brain and OpenAI, where they were exposed to the rapidly evolving machine learning field. Among their early mentors: Google’s Jeff Dean and OpenAI’s Sam Altman.
“We had this really interesting insight into a lot of early work around transformers, which ultimately is enabling a lot of the technologies that we’re seeing today,” said Buduma. “From 2017 to 2020, we just saw this inflection point of experiments that worked and, just as importantly, didn’t work…We also realized that these general models are likely to get much better in general reasoning tasks, like math and software engineering and copywriting. But there’s a big gap between the most capable general models and the most capable clinical models.”
I spoke to Dr. Richard Long, who’s a urologist at John Muir, about his experience using Ambience. He said that one of the invisible benefits of the technology is that it can ameliorate doctor burnout—a well-documented but relentless problem.
"I was getting to the point where I wasn’t sure I wanted to keep doing this,” said Long. “I love medicine, but I was trying to figure out ways that I could just spend my life in the operating room and do medicine without having to do all the documentation…But this is helping tremendously. The burnout is real, it’s personal.”
That said, it’s not unreasonable to wonder if the eventual progression of this technology is to replace doctors altogether. I raised this to Dr. Tanya Threewitt, VP of ambulatory health at John Muir, and she’s optimistic that the technology will serve as a complement rather than a substitute. "I don't know that it's ever going to be able to replace our knowledge,” she said. “It's definitely not going to replace the human touch, and I don't think you can say enough about how important that is."
Another big question is whether the market is large enough to sustain so many competing VC-backed AI products catering to doctors? Time will tell, but I’m unconvinced this is a winner-take-all space (there are, after all, lots of hospitals out there). And this seems to be an up-and-to-the-right story, which is quite a sight in healthcare, known for being spectacularly rigid and regulated.
And Ng says he’s getting encouraging feedback from one key demographic: “Since we’re in the Bay Area, one of the bonus things we get is that every now and then an investor or friend says, ‘oh my gosh, my clinician’s using Ambience.’”
See you tomorrow,
Allie Garfinkle
Twitter: @agarfinks
Email: alexandra.garfinkle@fortune.com
Submit a deal for the Term Sheet newsletter here.
Nina Ajemian curated the deals section of today’s newsletter.
