Arcadia CEO Michael Meucci on advancing AI and avoiding setbacks

Many healthcare organizations are struggling to find a sense of balance when it comes to incorporating artificial intelligence, Michael Meucci says.

Meucci, the president and CEO of Arcadia, a health data company, says providers and payers are looking to use AI in more ways, while at the same time being cautious about sending more data to more places safely, including guarding against cyberattacks. He likened the situation to pressing on the gas pedal and the brake pedal at the same time.

“You have AI, which we can, at this point kind of definitively say, will transform the way we provide care, whether it's the way that we access care as consumers, the way that providers gain insights on how to drive diagnostic journeys, the way that data is exchanged … it's going to impact everything,” he says. “How do you implement that at scale, safely, with compliance?”

Meucci spoke about advancing AI in an interview for Data Book, a podcast from Chief Healthcare Executive®. The new podcast dropped today.

• Read more: AI in health care: 26 leaders offer predictions for 2026

Meucci stresses the importance of developing AI safely, including making sure people are monitoring the effectiveness of those solutions. But he cautions against setting standards too high in evaluating AI tools in health systems.

“We have to remind ourselves that as humans, we're fallible, too,” he says. “We make mistakes all the time. And AI at this point is really a reflection of us. It's trained on evidence that we created and we've published. And so if AI is using the same body of evidence that we as humans are doing, it's going to be fallible too. And that's why we need to have systems where humans are in the loop.”

Meucci suggests that healthcare workforces are going to have to be trained on how to use AI tools. He says some new projects with AI tools can stall or have setbacks when “the change management effort was understated.”

“I think that that's where we need to mature overall, in how we're implementing these solutions is building not just the tech implementation, but also the change management,” he says.

Meucci recalls the early implementation of electronic health record systems, when organizations expected employees to use those systems and be as productive as they were before, but that didn’t exactly come to pass.

“We’re moving into that era again,” he says. “It's going to be AI optimization. How do you change your workflows, your staffing models, your patient education, your provider education, your technology and administrator education, to really consider how AI should be used?”

Meucci says it’s important to understand that these are the early days of using AI agents and large language models.

“It’s the worst AI we're going to use,” he says. “The AI models that we have today are enormously better than what we had a year ago, and it's still the worst technology we will use, because they just continue to improve.”

Healthcare organizations must develop strong governance systems for the development and deployment of AI. Meucci disagrees with entrepreneurs who say spending time on governance will slow down development.

“I take an opposite view. I think that an organization, whether you're a payer or a provider or a life sciences organization, that has a strong view on AI governance, can be a huge accelerator to adoption,” Meucci says.

Governance policies need to be living documents that are updated as AI technology evolves, he adds.

Organizations need to think about AI agents as employees, he suggests. Health systems utilize quality and safety committees to review errors by clinicians, and they need to take the same step for AI agents. And that feedback needs to be shared with the individuals building and training those agents, he says.

“I think that we need to, as a society, figure out how we're going to supervise AI, where we need to have a human in the loop, and where AI can't and isn't ready to perform. And those are tests that need to be continuously monitored and adjusted,” Meucci says.

Some AI applications can improve productivity and satisfaction without touching patient care, such as streamlining the process of prior authorization for treatments, billing, and coding.

Organizations need to ensure that governance models account for uses of AI that could lead to risks in delivering patient care. But for administrative cases, Meucci says health systems should look at “a faster lane to test and deploy AI.”

Check out the full conversation as Meucci weighs in on the uses of AI, the concerns of payers and providers, and the gap between strong and struggling health systems.

Here’s the full conversation. You can subscribe to Data Book wherever you get your podcasts.