The runaway hype of artificial intelligence has not diminished even as health system users struggle to find a return on investment.
In the first month of 2025 alone, investors poured sizable funding rounds into AI-focused startups, including Qventus ($105 million), Innovaccer ($275 million), Hippocratic AI ($140 million) and Eleos ($60 million). But experts say these investors may be valuing AI companies too generously if health systems can’t justify the costs of these tools without reimbursement.
Related: Digital health companies pitch employers on cutting GLP-1 costs
The problem was succinctly summed up last month at the J.P. Morgan Healthcare Conference in San Francisco by Tempus AI CEO Dr. Eric Lefkofsky: "We haven't yet really figured out how to pay for AI in healthcare.”
Many of the AI tools available will generate no return for users, said Erik Pupo, commercial health information technology advisory director at consulting firm Guidehouse. As a former chief information officer at an academic medical center, he advises health system clients to thoroughly understand the vendor landscape before choosing a solution.
“Is there a lot of hype, a lot of overvaluations, a lot of money being spent and a lot of overblown marketing claims with AI? The answer is yes,” Pupo said. “Health systems better have a good AI strategy. They better have good AI governance."
Most healthcare AI tools are not reimbursable. As a result, health systems are looking to see if AI can make clinicians more efficient, said Dr. Keith Dreyer, chief data scientist at Boston-based Mass General Brigham. In most cases, the technology isn’t adding an output that would justify the investment, he said.
Even in radiology, where AI investment is higher than any other clinical area, there is not always a clear path toward ROI, Dreyer said.
“Most AI approved by the [Food and Drug Administration] does a single function, such as looking for pulmonary nodules or breast cancer,” Dreyer said. “When you do that, the question becomes, 'Is the AI really helping the radiologists and making them more effective? Is it making them more accurate? Is it doing anything that justifies the purchase?' Many times, the answer is no.”
Area of optimism in AI documentation
Some AI tools, such as those used for documentation and capacity management, have developed a clearer sense of return. AI documentation has become a popular area of investment because it can reduce clinician burnout and give doctors more time with patients.
While these metrics don’t necessarily equate to a financial return, they’re easier to justify when clinical burnout is rampant and clinician shortages persist. Keeping these doctors happy is more than important than ever, Pupo said.
Still, some health systems are looking at time savings derived from ambient AI as an opportunity to increase productivity and receive a tangible financial return. Initially when safety-net system Denver Health added an ambient AI documentation tool from vendor Nabla, Associate Chief Medical Information Officer Dr. Daniel Kortsch said he asked all pilot users to increase their productivity by a very small margin so the investment could be net neutral.
After the pilot, the requirement was rolled back from two extra patients to one per month. But the main driver of the tool has been clinician happiness, Kortsch said.
“Burnout is at an all-time high right now, and we’re hoping this can make a change,” Kortsch said. “We had 83% of our pilot users say that Nabla would increase their desire to maintain clinical hours at Denver Health. That, in and of itself, is just really exciting."
But even these successful AI tools are only generating returns that affect a small percentage of the hospital, said Christopher Whelchel, a healthcare AI consultant. As AI spending goes up, he said vendors should expect more questions on how it can make a difference across the organization.
“If I walked into a hospital, pulled 10 people, and asked them, ‘How is AI helping you today?’ most of them would probably look at me cross-eyed and say, ‘I have no idea what you’re talking about,’” Whelchel said.
Investors plow ahead on AI
Despite the challenges with ROI and a lack of reimbursement, investors are not slowing down in their overall interest in AI. In 2024, 30% of investments in digital health were companies that use AI, according to a January report from Silicon Valley Bank.
Investors, cognizant of digital health’s history and the overvaluation of companies in the sector, are scrutinizing their investments more than ever, said Steve Kraus, a partner at venture capital firm Bessemer Venture Partners.
Kraus, who is an investor in Qventus, said the AI vendor has generated a significant return for its health system customers. Qventus’ success in this regard was critical in securing a follow-on investment from Bessemer, which first wrote a check for the company in 2018.
“There is hype and mania around anything AI right now,” Kraus said. “With Qventus, I’ve talked to the actual customers and have seen the value it's created.”
In some cases, health systems and vendors are willing to work together to experiment with pricing models that ensure there is a return on investment, said Prateesh Maheshwari, an investor at San Francisco-based Maverick Ventures. Vendors are also sometimes only charging customers based on usage of the AI tool, he said.
Even with higher levels of scrutiny, Maheshwari said adoption of AI among providers is still in its early innings and their interest in the technology hasn’t waned.
“I don't know how it will play out, but it feels like budgets are being set aside,” Maheshwari said. “Management attention is being focused on it.”
