Health systems moving AI initiatives from pilot programs into production are discovering that the technology's success depends less on model sophistication and more on the quality of the data underneath. As organizations prepare for broader deployment of agentic AI, ambient listening, and generative AI tools, many are confronting an uncomfortable truth: years of data investments have not produced environments capable of supporting enterprise-scale artificial intelligence.
"Many healthcare organizations are approaching AI as a software deployment when it is really a data infrastructure initiative," said Dr. Jaime Bland, CEO of Aquila Health.
From pilots to production
Proof-of-concept projects can mask underlying data problems because they typically run on carefully curated datasets assembled specifically for testing. Data teams clean records, reconcile inconsistencies, and prepare information before introducing it to the model. The results often look impressive. But production environments operate on live data feeds, and the difference is stark.
"Production, on the other hand, runs on the live feed, and the live feed is messy, non-standard, and incomplete in ways the pilot never surfaced, because it was never run on live data," Bland said. Clinical systems, workforce platforms, financial applications, and operational tools frequently maintain separate versions of the truth. Even within a single EHR, clinicians document information differently, creating inconsistencies that amplify when AI tools attempt analysis at scale.
"Most of the AI being deployed in health systems right now sits on top of that flawed, fragmented data, so the results it produces are poor," Bland said. For CIOs pursuing enterprise AI, the challenge shifts from technology procurement to data modernization-and the organizations seeing the greatest friction are not struggling because of the AI itself but because of the foundation beneath it. For those in leadership roles, an AI Learning Path for CIOs can help bridge the gap between strategic ambition and infrastructure readiness.
Interoperability solved movement, not quality
Federal policies, information-blocking regulations, and TEFCA have improved the industry's ability to exchange information. But Bland argues that interoperability addressed only part of the challenge. "Interoperability investment did what it set out to do. It made data move," she said. "What it did not do is make that data worth retrieving."
Patient information now flows between organizations more freely, yet assembling a complete, usable patient record remains difficult. Records arrive using different identifiers, coding systems, and documentation structures. Consent requirements vary. Reconciliation often demands significant manual effort. For rural and under-resourced organizations that lack staff to normalize incoming information, these barriers are particularly acute.
Governance gaps where they matter most
Healthcare leaders often frame AI governance around model oversight, transparency, and safety. Bland sees an equally pressing issue that receives less attention: data governance. When assessing AI readiness, she encounters the same problems repeatedly. Patient identities do not align across systems. Clinical records conflict with payer data. Outcomes information is incomplete. Social determinants data may be sparse. Race and ethnicity fields are frequently inconsistent.
"The gaps tend to sit where they matter most, since outcomes, social factors, and race and ethnicity are usually the thinnest fields, which is also where bias hides," Bland said. "AI cannot find a pattern that is not in the data." These shortcomings often remain invisible during testing because models are evaluated against datasets containing the same limitations present during training. The weaknesses emerge only after deployment, which means AI governance-for bias, reliability, and clinical trust-cannot be separated from data stewardship. This intersection is central to AI for Healthcare strategies that prioritize responsible implementation.
Why this matters for healthcare professionals
The next competitive divide may not fall between organizations using AI and those that are not. It may fall between those that invested in trustworthy data foundations and those that layered increasingly powerful AI tools on top of information they never fully fixed. Bland puts it simply: "The models will only get more capable from here, and that part keeps getting easier. The data underneath them is the slower, harder work, and it is what decides whether that capability turns into something a clinician can rely on."
For CIOs, clinical leaders, and operations teams, the implication is clear. Before betting on the next AI application, organizations should ask whether their data infrastructure can support it. That means treating data as a strategic product, assigning ownership, building governance structures that enable responsible innovation, and monitoring data quality continuously-work that generates fewer headlines but ultimately determines whether AI delivers on its promises in patient care.
Your membership also unlocks:
