In the past, healthcare AI startups were able to raise capital or secure pilots based on their potential and the credibility of their founders — but now, the bar is higher. Investors, as well as health system and payer customers, prefer startups that have demonstrated proven value, according to a panel of experts.
Investors and customers alike have become more skeptical about AI startups in the past couple of years, often demanding to see published research, case studies showing clear ROI and data on commercial traction before committing, said Nick Culbertson, managing director of Techstars, an accelerator launched in partnership with Johns Hopkins University and CareFirst BlueCross BlueShield. He made these comments during a panel discussion last month at MedCity News’ INVEST Digital Health conference in Dallas.
“A lot of hospital systems were saying, ‘Well, we want to be seen as innovative. We’re willing to spend and invest on this project and hope it pays off.’ I think over time, a lot of investors and a lot of health systems have been burned by companies that they gave a little bit too much leeway to and then it did not pan out,” Culbertson explained.
He said that AI is making the most immediate and meaningful impact in administrative and compliance workflows, noting that automating these back-office tasks can significantly reduce hospitals’ labor costs, as well as free up clinicians to focus more on patient care.
Dr. Ngoc-Anh Nguyen, vice chair of research at Houston Methodist’s innovation center, agreed that AI’s clearest value in healthcare so far is administrative rather than clinical.
She pointed out that physicians already know how to deliver care and most trust their own medical judgment over AI. In her view, they need AI to simplify administrative burdens and compliance tasks, not to make treatment decisions.
Dr. Nguyen also noted that physicians want polished, easy-to-use products.
“A physician is already always stretched to 110% for delivering patient care. The PCPs are getting scheduled for 10, 15 minutes with new patients. We’re seeing the patients, we’re documenting, then we’re having to be compliant — so the last thing we want is more work to learn to use another tool,” she declared.
If a tool has a burdensome interface or demonstrates poor accuracy, adoption at scale is impossible, especially among older physicians who are resistant to new technology, Dr. Nguyen added.
Another panelist — Eric Levine, principal at consulting firm Avalere Health — pointed out that the same scrutiny hospitals are applying to AI startups is also playing out among payers.
For payers, value can have very different definitions depending on the line of business, such as Medicare Advantage, Medicaid or commercial. For example, improving Star ratings, risk adjustment accuracy or reprocurement odds could matter as much as direct cost savings for a Medicare Advantage plan, Levine explained.
Overall, he noted that payers can be “a lot harder to crack” for AI startups.
“[Payers] can be very risk-averse in a lot of areas, and they really expect, two to three times ROI or they won’t even get in the door with you,” Levine remarked.
When trying to win over a payer, it’s crucial for startups to show evidence of their value — and that evidence must match the payer’s population, he noted. Many companies showcase data from studies on narrow or high-risk populations that don’t reflect a payer’s members, which undermines credibility.
The panelists agreed that the next wave of healthcare AI success stories won’t come from the flashiest models or largest funding rounds — but rather from the startups that can prove they work in the messy reality of patient care and payer contracts.
Photo: MedCity News
