Why healthcare CIOs must fix their foundations before scaling AI

Lukasz Lazewski, CEO of LLInformatics

The healthcare industry has entered a new phase of artificial intelligence adoption.

The early rush to experiment with generative AI has given way to a more difficult question facing CIOs and digital leaders: How do health systems move from isolated pilots to enterprise-scale AI that is secure, governable, clinically defensible and adaptable to rapid technological change?

That challenge arrives at a time when healthcare organizations are already grappling with aging infrastructure, cybersecurity threats, workforce shortages and mounting pressure to demonstrate measurable returns on technology investments. While vendors continue to promote increasingly powerful AI models, many provider organizations are discovering that the real bottlenecks lie elsewhere.

According to Lukasz Lazewski, CEO of health IT consultancy LLInformatics, healthcare's AI future may depend less on selecting the right model and more on fixing foundational problems that long predate generative AI.

"Most AI pilots in healthcare do not fail because of the wrong model or a flawed implementation – they fail because the organization simply is not ready," he said.

That observation reflects a growing reality across the industry. As health systems mature beyond experimentation, CIOs are increasingly confronting questions about data readiness, governance accountability, model oversight, interoperability and long-term sustainability.

The next wave of AI adoption may ultimately be won, not by organizations deploying the most AI tools, but by those building the strongest operational foundations underneath them.

Healthcare's biggest AI obstacle

Healthcare executives often describe data as their most valuable strategic asset. Yet in practice, clinical, operational and financial information remains scattered across EHRs, departmental systems and third-party applications.

For AI initiatives, that fragmentation creates significant challenges.

"The first is data. Health systems store data across multiple systems with no proper aggregation, cleanup or normalization," Lazewski said.

The problem becomes more acute inside legacy environments. Many health systems continue to operate highly customized EHR deployments built over decades of acquisitions, mergers and incremental technology investments. Those environments may function adequately for transactional workflows but often create obstacles for modern AI architectures.

"Without solving that first, you are running an experiment on top of a broken foundation," he said.

The issue extends beyond technology. Enterprise AI requires organizational alignment across clinical, operational and administrative stakeholders. Pilots frequently succeed within a single department only to stall when broader adoption requires cooperation across the enterprise.

Lazewski noted that initiatives can become trapped by competing priorities, limited engagement and conflicting incentives among leadership teams.

Governance is becoming a board-level issue

As AI begins influencing clinical workflows, patient communications and operational decision-making, governance is rapidly moving from a technical discussion to a strategic one.

Many healthcare organizations still approach governance reactively, establishing oversight processes only after deployment challenges emerge.

"The questions that should be answered before a pilot launches, including who approves the model, who monitors it, who is accountable when it produces a wrong output, and how that is tracked and corrected, are typically only raised once problems emerge," Lazewski said.

Those questions are becoming increasingly important as regulators, accreditation organizations and legal experts begin examining how AI-generated outputs should be validated, monitored and documented.

For CIOs, the issue is no longer simply whether AI can produce useful outputs. The question is whether organizations can consistently explain, audit and defend those outputs.

Lazewski argues that monitoring, traceability and accountability must be built into AI programs from the beginning rather than retrofitted later.

He also warns that organizations often overlook life cycle management.

"Software has a life cycle. It is built, it matures, it ages, and eventually it sunsets, and at every stage of that life cycle, you need people in place to look after it and ensure business continuity," he said.

Future-proofing in an era of rapid AI change

Healthcare leaders face another challenge that is unusual even by technology standards: the pace of AI innovation.

Models that were considered cutting-edge two years ago have already been eclipsed. Many experts expect the current generation of large language models to undergo similar disruption during the next several years.

That reality creates significant architectural implications.

"The single most important quality of a future-proof AI architecture is how plug-and-play it is," Lazewski said.

Rather than tightly coupling business processes to specific models, he advocates modular architectures that separate data, integrations, AI services and governance functions.

"In concrete terms, this means keeping a clear separation between your data layer, your integration layer that covers both internal and external tooling, your AI model layer, and your governance and monitoring layer," he said.

The approach reflects a broader trend emerging among large health systems. Technology leaders increasingly view AI as another enterprise capability layer rather than a standalone application.

That distinction matters because future AI strategies may involve continuously swapping models, retraining workflows and introducing new capabilities without disrupting core clinical operations.

According to Lazewski, many organizations continue to approach AI incorrectly.

"They treat AI as an IT project with a defined endpoint rather than as a transformation of the entire organization, its workflows, its culture and its people," he said.

Technical debt threatens AI return on investment

Perhaps no issue looms larger for healthcare CIOs than technical debt.

Years of deferred modernization, undocumented integrations and legacy code create hidden costs that frequently surface during AI deployments.

"Technical debt is probably the single biggest hidden cost in healthcare AI," Lazewski said.

The consequences appear throughout implementation efforts. Integration timelines lengthen. Data remediation consumes budgets. Development teams spend increasing amounts of time fixing infrastructure problems rather than delivering new capabilities.

"If your data is clean, your systems are well-integrated, your APIs work in concert, and your existing workflows are performing their business objectives well, then AI can genuinely accelerate things," he said. "If they are not, AI does not fix them. It scales whatever is broken."

Technical debt also creates cybersecurity and compliance concerns. Older systems can be more difficult to monitor, harder to secure, and less capable of supporting the auditability increasingly expected of AI-enabled environments.

"If a regulatory body asks you to demonstrate how your AI reached a particular decision, and your underlying architecture is built on top of technical debt, you may find yourself unable to answer," Lazewski said. "That is not a compliance risk. That is a compliance failure."

The strategic takeaway for CIOs

The healthcare industry has spent the past several years asking what AI can do. The next several years may be defined by a different question: What organizational conditions must exist before AI can safely scale?

Lazewski believes successful organizations are already moving in that direction by treating technical debt remediation, data governance and architectural modernization as prerequisites rather than side projects.

He advocates incremental modernization efforts that produce measurable wins while gradually improving enterprise readiness.

Ultimately, he sees successful healthcare AI programs resting on three pillars: trusted data, flexible architecture and strong governance.

"Every successful healthcare AI program I have seen rests on three things: strong data that is high-quality, accessible and well-governed; strong architecture that is modular, secure and interoperable; and strong governance," he said.

For healthcare CIOs facing growing pressure to demonstrate AI value, that may be the most important lesson. The AI model itself is often the easiest part. The harder challenge is preparing the organization around it.

Follow Bill's health IT coverage on LinkedIn: Bill SiwickiEmail him: [email protected]Healthcare IT News is a HIMSS Media publication.

WATCH NOW: Emerging advancements in radiology AI