By Sandra Johnson, CDH-E, SVP of Client Services

All eyes in healthcare are on AI, but beneath the excitement lies a difficult truth: intelligence built on fragmented data can’t deliver reliable results. For CIOs— who’ve watched their roles in directing and deciding AI strategies swell from 31 to 86 percent over the past year— more pilots and promises are not the answer. Instead, they want clarity, connection, and confidence that the insights they’re receiving are ones they can rely on. That’s a tall order, and until interoperability catches up, AI will continue to be a noisy experiment rather than a trust ally for delivering quality care.

AI has become an experimentation trap

Healthcare CIOs are eager to adopt and reap the full benefits of AI, and they’re trying, but industrywide adoption is taking time. According to the 2025 CHIME Survey, around 52% of healthcare organizations say they’re already actively using AI-powered tools, and the other 48% are evaluating them for future use. Even still, only a fraction— around 2%, according to a recent report in Becker’s— describe AI tools as advanced. Leaders are actively experimenting without full infrastructure, left to evaluate too many tools with too little time. As a result, most are seeing the same patterns, like AI projects that create more complexity than value. But CIOs aren’t looking for another vendor promise. They instead want a way to make innovation sustainable, measurable, and safe, and AI isn’t getting there on its own.

Embedded AI isn’t the ‘next’ imperative, it’s now

For AI to truly scale, it has to live where care happens: inside the electronic health record itself. The CHIME survey found that 48% of CIOs believe it is extremely important for AI capabilities to be natively embedded within their EHR, and another 44% believe it is somewhat important. The message from CIOs is unmistakable, they’re done stitching together third-party apps and one-off integrations that add more governance work and risk without all of the benefits of a single, integrated patient record. They want AI that’s natively embedded, secure by design, and compliant by default without all of the additional work. Embedded intelligence doesn’t just streamline adoption. It also transforms AI from being just another external experiment into a built-in, powerhouse capability. And when AI becomes inherent to clinical workflows and EHRs, health leaders can move past pilot fatigue and on to the high performance and quality of care they were seeking all along.

Interoperability is the engine of healthcare intelligence

But the truth is that embedded AI is only as good as the data that powers it. According to the CHIME survey, 100% of respondents said AI-driven interoperability is crucial for improving data sharing and care coordination. Another 32% believe interoperability is extremely important and 68% say it’s at least somewhat important. Despite this, just 26% of respondents listed increasing interoperability and data accessibility as a top AI priority, signaling a gap between recognition and real-world action. What remains clear is that without interoperability, hospital systems can’t see the full patient story. Equally as important, AI can’t learn from data that it doesn’t have access to. Interoperability isn’t a side project that can be completed any time down the line. It’s a precondition for trustworthy and reliable health intelligence from an EHR. CIOs need one source of truth— just as clinicians need one integrated patient record— not ten disconnected signals from multiple different third-party apps. And when clean, connected data becomes the default, AI can amplify insights instead of producing more noise.

Five key takeaways for CIOs

If CIOs have made just one thing abundantly clear, it’s this: AI innovation can’t advance faster than the infrastructure that supports it, not if the goal is trustworthy, reliable insights. CIOs who want to move on from experimentation to real execution can start with five practical steps:

1. Interoperability— it’s the foundation of any serious AI investment.

2. Treat native integration as a requirement, not an optional upgrade— tools that live within your EHR will reduce friction, frustration, and risk.

3. Simplify your governance by limiting the number of disconnected, third-party apps— choose systems that manage AI transparently and securely within your already existing compliance frameworks.

4. Prioritize solutions that deliver meaningful usability and functionality — not features that look impressive but don’t advance your clinical, operational, or financial goals.

5. Build for the future of a single, integrated patient record— and make unified, normalized data the foundation for better care and more reliable intelligence.

The last thing healthcare needs is more hype-driven AI. Instead, what providers need is a path forward that’s focused more on interoperability and functionality than on getting things done quickly. When AI is built on a foundation of true interoperability, it stops being noise and starts becoming knowledge — the kind that empowers clinicians, strengthens operations, and transforms care.

Link to the original article.