Health systems should use data to identify problems before buying AI tools, according to Arcadia CEO Michael Meucci.
The market is flooded with vendors pitching AI solutions — but instead of responding to these pitches, Meucci thinks health systems should analyze their own data to find inefficiencies, then seek AI tools that address those specific issues.
“What I keep encouraging my customers to do is use the data they have to tell them, ‘Hey, you’ve got clinical variation over here. You should go look for a solution to reduce that because the ROI opportunity could be X.’ It’s changing from an inbound demand to a data-driven project,” he stated during an interview earlier this month at the HIMSS conference in Las Vegas.
Meucci said this approach helps organizations to benchmark outcomes and properly attribute ROI among different AI tools.
Financial pressures are forcing providers to more rigorously measure the value of AI solutions — and without strong measurement frameworks, he said it’s difficult to determine which vendor actually generated improvements.
As health systems become more disciplined about evaluating their AI projects, they’re also expanding how they define ROI, Meucci noted.
There are many AI tools that may not produce obvious financial returns like increasing visit volume, but they still generate meaningful ROI — such as reducing administrative burden and improving physician satisfaction and retention, he explained.
For example, ambient documentation tools — like those sold by Microsoft, Abridge, Suki and others — have done a good job of reducing physicians’ documentation burden, which eases burnout and makes clinicians more likely to stay in their roles. At scale, this trend can generate serious cost savings for a health system, given that replacing a physician can cost about $1 million, Meucci pointed out.
In his view, AI is becoming part of the physician experience strategy — and he thinks hospitals will soon begin promoting their technology stack as a recruiting and retention tool.
He drew a parallel with the concept of “developer experience” in software, where companies equip their engineers with the tools they need to work more efficiently. Meucci said health systems are starting to think similarly about clinicians, evaluating which technologies can decrease their workload and make it easier for them to focus on patients.
“I was talking to the CFO of a pretty large health system in Massachusetts, and he said they now have a work group to think about the physician experience. They ask, what ambient tools do we have available? What computer-assisted coding tools do we have available? How do we help reduce the administrative burden?” he remarked.
Meucci also noted that switching between AI vendors is becoming easier as interoperability improves. Open APIs, standardized data formats and more unified data infrastructures are allowing health systems to deploy new tools quickly without getting locked into one platform.
This means hospitals can pilot different solutions, compare results, and then replace a tool if it doesn’t deliver a good enough ROI. Over time, Meucci thinks these lower switching costs could encourage more experimentation and faster adoption of the most effective AI models.
Together, these trends are pushing health systems toward a more disciplined approach to AI adoption — one driven by data and measurable outcomes.
Photo: Issarawat Tattong, Getty Images
