In the past year-plus, Epic has ended its startup codevelopment program and rolled out its own ambient AI scribe to compete directly with the companies that helped popularize the category in the first place.
It’s a live example of the question a panel of health system leaders and startup operators wrestled with Sept. 23 at Mass General Brigham’s World Medical Innovation Forum in Boston: Once an incumbent platform like Epic or Microsoft decides it wants what a smaller company built, what actually keeps that startup alive?
The panel, moderated by healthcare investor Brenton Fargnoli, MD, drew a mostly consistent answer: A startup survives by owning something the platform can’t easily absorb.
Ada Glover, chief product officer at interoperability startup Zus Health, said the clearest tests are which capabilities are central to the EHR.
“Those AI scribes are going after the most core element of the EMR, arguably, which is documentation,” she said. “That’s clearly going to be an area where the EMRs themselves really dig in and try to establish a footprint.” Radiology and imaging, by contrast, are farther from that foundation, she said, leaving startups more room to build a defensible position.
Ms. Glover said the second risk factor is how dependent a startup is on a single EHR. A tool that only reads from one system and hands results back into that system’s own workflow is vulnerable once the platform builds a competing feature natively, she said. Zus has tried to avoid that by building what Ms. Glover called “an interoperability platform” that relies not just on a single customer’s EHR, but on “an entire ecosystem of data services.”
Terrence Chen, MD, CEO of AI company United Imaging Intelligence America, framed defensibility even more directly around information. Companies that produce their own clinical data — through imaging systems, surgical robots or wearable devices, in his company’s case — have an inherent leg up.
“You no longer just consume data, but you also generate data,” he said. Without that advantage, he said, a startup needs another moat: proprietary data and labels, deep domain expertise in a specific clinical workflow, or evidence built up over years of use. The closer a company sits to where clinical action actually happens, he said, the harder it becomes for a larger platform to simply build a copy.
William Gordon, MD, a Mass General Brigham hospitalist and senior advisor at law and policy firm Manatt Health, offered a different dividing line: whether a startup is willing to take on accountability for an outcome that an enterprise platform won’t. Task- and process-oriented tools, he said, are harder to disrupt because incumbent systems already own that territory.
“It’s unlikely that an EHR vendor is going to say, ‘We are actually going to be accountable for our patient’s blood pressure improving,'” he said. A newer entrant, by contrast, might tie its own payment to the outcome itself — a level of risk, and a value proposition, he said an enterprise system “just doesn’t need to and won’t take.”
Eric Podradchik, vice president of digital clinical systems at Somerville, Mass.-based Mass General Brigham, said startups that operate independently of a health system’s existing workflows can work for a while but eventually hit a ceiling.
“Without the integration, it becomes much more challenging to expand to a broader use case,” he said, adding that health systems increasingly evaluate vendors first on how closely they align with their core platform’s own roadmap before considering anything built outside it.
Ms. Glover put a finer point on what all of that adds up to financially: A startup can survive on goodwill and improved quality of life for clinicians for only so long. “We’re capitalists here,” she said. “There needs to be a path to green money, not just blue money” — a return on investment clear enough that token budgets and staff time can keep being justified, not just a nicer workflow.
