For nearly two decades, healthcare's digital transformation efforts have centered on a single objective: getting data to move. Federal incentives, electronic health records, health information exchanges and interoperability standards have dramatically improved the industry's ability to share patient information across organizations.

But as artificial intelligence rapidly moves from isolated pilots to enterprise-scale deployments, healthcare leaders are discovering that connected data alone isn't enough. Increasingly, the conversation has shifted from interoperability to something far more ambitious: data utility.

That's the premise Julie Smith, head of global healthcare market strategy at InterSystems, and Jonathan Teich, M.D., Ph.D., chief medical officer and director of clinical innovation at InterSystems, explored during a recent webinar hosted by the American Medical Informatics Association (AMIA). Their message reflected a broader reality emerging across healthcare IT: the organizations that will realize AI's full potential won't necessarily be those with the most data. They'll be the ones with data that AI can actually trust.

As Teich explained, "Interoperability helps us; it helps get us the data we need, but it's not the total finish line. It's really the starting point for something that's a lot more demanding."

AI Has Changed the Rules

Healthcare organizations have spent years investing in interoperability. Most large health systems today can exchange clinical information through Fast Healthcare Interoperability Resource (FHIR) application programming interfaces (API), health information exchanges (HIE) and national networks. That progress remains essential.

Yet AI introduces entirely new requirements. Large language models, predictive algorithms and emerging agentic AI systems don't simply retrieve information. They interpret it, synthesize it, make recommendations and increasingly initiate actions. Every weakness in the underlying data becomes amplified.

"AI puts new pressure on any kind of weakness," Teich said. "Gaps in completeness, gaps in accuracy, meaning, provenance, timeliness — anything that doesn't work quite right here is magnified through AI."

In other words, AI doesn't expose data problems, it magnifies them. Incomplete documentation, inconsistent terminology, duplicate records and missing clinical context may have been manageable when clinicians manually reviewed patient charts. They become significantly more consequential when AI models are expected to generate reliable insights at enterprise scale.

"Interoperability got the data moving, but the utility is what gets the data functional."

— Jonathan Teich, M.D., Ph.D.

From Data Exchange to Data Utility

The webinar repeatedly returned to one concept that deserves far more attention: data utility.

Unlike interoperability, which focuses on moving information between systems, data utility emphasizes making that information immediately usable for care delivery, operational decision-making, research and continuous improvement.

That requires more than simply collecting data. Organizations must normalize information from multiple sources, preserve clinical context, document data provenance, establish governance policies, maintain semantic consistency and integrate insights directly into clinical workflows.

Perhaps most importantly, clinicians have to trust what AI produces. "Trust is really what makes adoption or breaks it," Teich said. "If clinicians don't trust the data, they're not going to use it."

That observation speaks to one of healthcare's biggest AI challenges today. Much of the industry's attention has focused on selecting models, evaluating vendors and identifying high-value use cases. Comparatively less attention has been paid to whether the underlying data foundation is mature enough to support those ambitions.

Data Strategy Is Becoming AI Strategy

Smith noted that healthcare leaders themselves are recognizing this shift. Rather than asking how to exchange more information, organizations increasingly want to know how data is standardized, prepared for analytics and delivered in ways AI can reliably consume.

Research cited during the webinar found that 86% of healthcare leaders believe preparation and readiness represent where interoperability creates the greatest value for AI-enabled organizations.

Smith pointed to a growing consensus among industry leaders that AI strategy cannot exist independently of data strategy. As she noted, healthcare leaders increasingly recognize that "you cannot have an AI strategy without a data strategy, and you cannot have a data strategy without an interoperability strategy."

That hierarchy illustrates how industry priorities have evolved. Interoperability remains foundational, but it is no longer the destination.

Building a Learning Health System

Perhaps the webinar's most compelling idea centered on the learning health system itself. Healthcare has long envisioned an environment where data generated during care continuously improves future care. AI has the potential to accelerate that cycle dramatically — but only if organizations can trust every stage of the process.

As Teich described it, health systems must move beyond simply collecting information and instead "close this loop: from data to insight to action to outcomes back to learning."

•        Each stage depends on the integrity of the one before it.
•        Poor-quality data produces unreliable insights.
•        Unreliable insights produce ineffective interventions.

Without measuring outcomes and feeding those results back into the system, organizations never truly learn.

That closed-loop approach represents one of the defining characteristics separating AI-ready organizations from those still experimenting with isolated pilots.