Health system leaders are racing to deploy artificial intelligence, but Tomas Gregorio, senior vice president and chief digital information officer at Care New England Health System in Providence, R.I., believes the industry is outrunning its ability to govern it.
Care New England completed its Epic implementation last October, a $100 million project serving as the foundation for AI deployment. Ambient voice documentation, automated coding support and denial management automation are already running on top of it. But as those administrative applications scale, Mr. Gregorio has maintained a deliberate boundary around clinical AI, and his reasoning points to a gap he says is widening between what the models can do and what the frameworks governing them can handle.
“We want to be at the crest of the wave, not at the bleeding edge of AI,” he said during an episode of the “Becker’s Healthcare Podcast.” “We’ve taken the approach that administratively, we want to reduce that burden as an organization. But the models are improving faster than the governance frameworks and I think within three to five years, we’re going to face real hard questions about AI’s role in diagnosis and treatment planning that we’re not fully prepared for as an industry. We’re not ready for that.”
The concern is structural. Most health systems are still building the frameworks that would define where AI can inform a clinical decision, where it requires mandatory human review and where it should not be deployed at all. The AI models those organizations are currently piloting have already outpaced the rules meant to govern them.
“I’m watching the frontier of AI models very closely, reasoning capable systems that can handle genuinely complex, multivariable problems,” said Mr. Gregorio. “Healthcare is full of those types of problems, and the organization that figures out how to safely deploy that capability will have a durable advantage. That’s not hype. That’s just the trajectory of the curve.”
Mr. Gregorio sees a parallel failure mode operating from the other direction. Organizations still running proofs of concept will be lapped by those treating AI as infrastructure. The path between moving too slowly and moving without sufficient governance is narrower than most leaders recognize, and the window to find it is roughly the same three to five years before the clinical governance question can no longer be deferred.
Cybersecurity is another big consideration. Hospitals are often targets of cyberattacks and AI has ushered in a new era of sophisticated attacks. President Donald Trump signed an executive order to strengthen AI-enabled cybersecurity on June 2 and has a close eye on frontier models. Anthropic shut off access to its frontier models Fable 5 and Mythos 5 after the federal government requested removing access from foreign governments, companies and other individuals. Earlier this year, Anthropic gave several companies access to Mythos, a large language model designed to supercharge computer security, to preview for defensive security work. Health systems and healthcare providers were missing from the group.
“We’re in the middle of a CISO build so that we can figure out how we can protect ourselves with AI itself,” said Mr. Gregorio. “Our partner is helping us go through that process. We have a big meeting as well with Anthropic and a lot of CISOs in the region. As an industry, we aren’t represented well as part of Project Glasswing and all of those initiatives that Anthropic has put together, which worries me because we don’t have the capacity to patch on demand at our systems. We have lives at stake and its a fiduciary quality issue for our board.”
In the next year, Mr. Gregorio expects AI to shift from pilot to operating assumption. Having AI built into the foundational EHR will give Care New England an advantage.
“The organizations that are still doing proof of concepts will be lapped by ones treating AI as infrastructure,” he said. “You need to get past that pilot stage at this point and that’s why we’re bullish about finding the right company to work with to make sure that we can be successful and move the needle slowly as we move forward.”
At the Becker's 11th Annual IT + Revenue Cycle Conference: The Future of AI & Digital Health, taking place September 14–17 in Chicago, healthcare executives and digital leaders from across the country will come together to explore how AI, interoperability, cybersecurity, and revenue cycle innovation are transforming care delivery, strengthening financial performance, and driving the next era of digital health. Apply for complimentary registration now.