Healthcare organizations are moving rapidly to adopt AI-driven tools, more than twice as fast as other industries, according to some reports. But some experts are concerned that the rapid pace of adoption won’t produce the clinical innovation the healthcare industry is seeking unless healthcare providers place a higher priority on data governance strategies.
“In my experience, I’ve seen a lot of AI pilots launched with a high level of excitement, only to stall in silence before achieving the innovations they promised,” says Chris Hutchins, founder and CEO of Hutchins Data Strategy Consulting. “In most cases, the breakdown isn’t caused by the AI algorithm. Rather, it’s the architectural framework beneath it. The culprit is typically a problem hidden in the data structure of a healthcare system.”
Hutchins Data Strategy Consultants was founded on a straightforward premise: healthcare organizations deserve data and AI partners who understand how healthcare actually works. Through team empowerment, Hutchins helps healthcare organizations enhance care delivery while reducing administrative work and transforming data into meaningful outcomes. His areas of expertise include enterprise data governance, responsible AI adoption, and self-service analytics.
“Too few healthcare organizations are focused on the data maturity needed to scale AI effectively,” Hutchins asserts. “They’re talking about ambient listening and clinical decision support when they should be focused on data quality, data lineage, and data protection. Without a solid governance program, data will be fragmented and inconsistent, which will keep AI tools from delivering on their promises while also creating operational risks.”
Healthcare data governance must establish a framework for ownership
For data governance to be effective in the age of AI, governance platforms must establish ownership. Governance efforts that only evaluate risk and prescribe policies won’t protect against the problems AI can introduce.
“Healthcare organizations follow a predictable pattern when it comes to addressing risk,” Hutchins says. “They form a committee, write a charter, and schedule meetings. That structure no longer works because it hinges on deliberation, whereas AI systems operate through execution. It’s a mismatch that must be addressed because it causes the accountability for patient data management and other essential governance functions to disappear.”
Hutchins points to three key questions that can be used to assess the effectiveness of a data governance program:
- Who approved this deployment?
- What performance thresholds were validated before it went live?
- What executive accepted the risk on behalf of the organization?
If the answers to those questions don’t establish ownership, organizations are leaving systems that influence care, documentation, and clinical prioritization vulnerable to adverse events.
“The failure of governance processes to establish an owner for AI oversight is a structural one,” Hutchins says. “And it doesn’t stay invisible. It surfaces when something goes wrong and stakeholders reach for governance records, finding only deliberation where accountability should have been established.”
How to implement data governance in the AI age
Deciding whether or not to deploy a new tool is an important phase of the data governance process. But in the age of AI, it can’t be the final phase. An effective governance program must establish a structure that remains in place after deployment, beginning by naming an owner and defining their authority and responsibilities.
“A governance structure can only protect the organization if it identifies the person responsible for every AI system influencing clinical decisions,” Hutchins warns. “Without taking that step, you’re just creating a record that shows you thoroughly reviewed a risk without assigning anyone to oversee it.”
Effective governance must also establish accountability, often requiring more than the audit trails AI platforms typically provide. Audit logs must go beyond simply showing that a program ran.
AI-guided decisions need more than a sign-off from a clinician. When a problem arises, the organization needs to be able to justify the decision. Records and accountability are not the same thing.
“For a clinician to be in control of a system, they need to know what information the AI model used to make its recommendation, as well as what the model skipped and what it was uncertain about,” Hutchins says. “When someone asks you to explain an AI decision — a patient, a regulator, a lawyer — you must be able to do it. If you can’t give a step-by-step explanation, your governance has failed.”
Healthcare organizations must act now to establish effective governance programs
The risks associated with faulty governance are not static. Organizations that continue to integrate AI without addressing holes in governance systems assume governance debt that compounds just like technical debt. The debt accumulates when ownership is unclear and responsibilities are deferred, spreading problems across systems and locking in exposure that can later become very costly.
“The cost of deferring governance is not linear,” Hutchins says. “Every AI deployment that goes live without proper oversight, accountability structures, and audit infrastructure adds to liability that becomes exponentially harder to address later.”
