AI’s Future in Healthcare: Balancing Innovation, Regulation, and Practicality

The rapid evolution of AI is reshaping healthcare, but navigating its complexities requires a blend of pragmatism and innovation, according to Dr. John Halamka, President of Mayo Clinic Platform. As health systems grapple with AI integration, the challenge lies in balancing cutting-edge advancements with regulatory constraints and operational realities.

Scroll down to watch or listen to the full interview; or subscribe to healthsystemCIO on your favorite podcasting channel

Having recently returned from discussions at J.P. Morgan and Davos, Halamka highlighted the uncertainty surrounding AI’s trajectory. “There’s a lot of contradictory information,” he said. “We see major market shifts driven by generative AI, but the key question is: where is it all leading?”

The AI Dilemma: Best-of-Breed vs. Enterprise Suites

For CIOs, the tension between adopting specialized AI solutions and sticking with enterprise suites remains a pressing issue. While niche vendors promise innovative capabilities, integrating thousands of individual products is unsustainable. “CIOs don’t want to manage 1,000 niche AI products,” Halamka stated. “Instead, they want AI that seamlessly fits within existing enterprise platforms.”

Yet, AI’s effectiveness depends on its applicability to diverse patient populations. Halamka pointed to the importance of AI models being developed with representative data sets. “An algorithm trained on Scandinavian Lutherans in Minnesota may not perform well in a New Jersey patient population,” he noted. The solution, he suggested, is a future where EHRs function like an app store—curating and deploying the best AI tools dynamically, rather than forcing health systems to commit to static solutions.

AI Strategy by Health System Size

AI adoption strategies should align with an organization’s size and capabilities.



* Large academic medical centers have the resources to develop proprietary AI models, leveraging vast datasets and computational power. At Mayo Clinic, Halamka’s team builds foundation models weekly. “With 250 predictive algorithms and eight foundation models, we have the infrastructure to innovate at scale,” he said.

* Mid-sized health systems benefit from acquiring commercial AI products and fine-tuning them with local data. “These organizations should focus on adapting proven AI models to their patient populations,” he advised.

* Smaller health systems should prioritize partnerships with vendors or academic institutions. “Collaboration is key—there’s no need for a rural hospital to invest in massive AI infrastructure when they can leverage expertise from larger institutions,” he explained.



The CIO’s Role: Leader or Order-Taker?

AI adoption should not be a reactive process. “No chief medical officer ever says, ‘I need more AI,’” Halamka noted. Instead, CIOs must align AI initiatives with organizational goals. “AI should solve real business and clinical problems—whether it’s reducing documentation burdens through ambient listening or improving referral workflows,” he said.

Halamka emphasized the importance of governance and strategic planning. “CIOs should not just take orders. They must help define how AI can accelerate key performance indicators,” he asserted.

The EHR of the Future: Evolution or Revolution?

Discussing the EHR landscape, Halamka acknowledged Epic’s dominance and the evolving role of Oracle Health. However, he questioned whether today’s EHR systems are designed for modern healthcare challenges. “If we started from scratch, would we build what we have today? Probably not.”

While a total overhaul isn’t feasible under current regulations, layering AI-driven interfaces onto existing EHRs is a logical next step. “We’re moving toward a world where AI abstracts away administrative burdens, allowing clinicians to focus on patient care,” he said.

Data and Algorithmic Integrity