Health systems have no shortage of AI startups pitching them. Many of those companies build on the same frontier models that EHR vendors, big tech companies, and health systems’ own data science teams can use, leaving technology leaders weighing a question beyond whether a product works: whether the company behind it will still matter in a year or two.
Three health system technology leaders addressed that question at Becker’s 14th Annual CEO + CFO Roundtable this month in Chicago, where Mark Cuban also used his keynote to warn health tech founders about competing with Epic.
Jeff Gautney, senior vice president and CIO of Chicago-based Rush University System for Health, spoke during a panel discussion on AI in the C-suite. Epic rolled out AI Charting, its own ambient documentation tool, in February, putting it in direct competition with ambient AI companies its customers already use.
“As the Epic tool gets better and better, it’ll probably be like Teladoc was,” Mr. Gautney said. “All of us were using Teladoc in 2021. Very few of us who are Epic customers continue to use an external video tool for our care. We use the Epic tool that’s embedded.
“And so everything that I buy now that’s a point solution like that, I assume it’s going to go away at some point and that I’m going to have to build it into my bigger chassis and increase my cost that way.”
At Atlanta-based Emory Healthcare, the EHR vendor is also the first stop when a department asks for an AI tool.
“If we have organic demand from a department or a unit or a leader for something, Epic is one of the first places we’ll check first to see if it’s already available or on roadmap,” Emory Chief AI Officer Nabile Safdar, MD, told Becker’s.
That doesn’t rule startups out. Avoiding them altogether would mean giving up innovative opportunities, Dr. Safdar said. But established vendors are watching those same companies. If Emory expects one of its vendor partners to offer what a startup does today within about 15 months, he said, it isn’t worth starting with the startup, especially when Emory sees nothing that sets it apart.
“There’s no vertical domain expertise, no special workflow sauce,” Dr. Safdar said.
Lisa Stump, executive vice president and chief digital information officer of New York City-based Mount Sinai Health System, expects most startups to have a brief window. With frontier models now largely interchangeable, she said, what sets a product apart is the agent and the user interface, and competitors can copy those quickly.
“So I think there’s going to be a short runway where if you’re first or early to market, you can make some money, and then I think it’s going to get diluted … as more and more get there,” Ms. Stump told Becker’s.
For Dr. Safdar, the test is who is on the startup’s team. A company led mainly by technology and business people, without staff who have worked in health systems and understand their workflows, is a turnoff, he said, because it signals Emory would end up training the vendor rather than getting solutions from it. Startups with deep experience in the problem they’re addressing can stay ahead of EHR vendors and other large companies for a long time, he said, or work in a niche that bigger players will never pursue.
“They can’t just be reselling a frontier model with a really good system prompt. It’s got to be deeper than that,” Dr. Safdar said.
Ms. Stump pointed to Latent Health, an AI company focused on pharmacy, as one point solution that clears that bar. Its pharmacy focus sets it apart, she said, because pharmacy remains the biggest contributor to the bottom line of every health system in the country.
Pricing is shifting the calculation, too. Buying can make sense when speed to value matters or when her team needs to focus on harder problems, Ms. Stump said, particularly as more AI companies agree to go at risk on their products, an arrangement she increasingly expects. Dr. Safdar said Emory is willing to work with startups that show they’re invested in its success and willing to take on that risk.
Mount Sinai also sets terms on data. Ms. Stump said the health system has turned down the model of vendors asking for all of its EHR and ERP data to run through their own tools, opting instead for a federated approach that gives vendors access to Mount Sinai’s data layer.
“The data layer has to stay ours,” she said.
She acknowledged that Mount Sinai’s sizable investment in its own AI scientists is rare. Organizations without that capacity will buy what they need or hire consultants to build in their own ecosystems, she said.
Mr. Gautney said he doesn’t see vendors as misleading buyers so much as selling to them, and consultants are now pitching AI solutions on contingency fees as well.
“So it’s buyer beware, as it always has been and don’t believe all the hype,” he said.
