Artificial intelligence is moving from the margins of healthcare into everyday practice. It can summarize medical research, draft clinical documentation, support diagnosis, improve imaging, identify patients at risk and reduce administrative burden.
The pace of adoption is striking. The American Medical Association’s 2026 survey found that 81% of responding physicians use AI professionally—more than double the 2023 share. Documentation and research summarization are among the most common uses. By September 2026, the U.S. Food and Drug Administration reported that it had authorized more than 1,600 AI-enabled medical devices across a wide range of specialties.
Figure 1. The pace of adoption. Sources: American Medical Association 2026 survey; U.S. Food and Drug Administration.
These developments create genuine opportunity. AI can help clinicians find relevant information faster, automate repetitive work, extend specialist expertise and identify patterns that people may miss. Used well, it can give health professionals more time for the work only people can do: listen, exercise judgment, explain options and care for patients.
But adoption alone is not transformation. The central question for healthcare leaders is no longer whether AI will enter the organization. It is whether the organization can demonstrate that each system is appropriate, secure, fair, understandable and accountable throughout its life cycle.
Why governance matters
Healthcare decisions are different from routine consumer transactions. An unreliable recommendation can affect diagnosis, treatment, outcome, access to services or the accuracy of a medical record. A model that performs well on average may perform less well for a particular language, age group, disability or demographic population. A system may also change through updates, new data or altered workflows after its initial review.
Governance turns these risks into defined management responsibilities. It identifies where AI is being used, who owns each system, what data it depends on, how performance is measured, when a human must intervene and what happens when something goes wrong.
Figure 2. Governance turns risks into defined management responsibilities.
Public-sector and international guidance increasingly point in this direction. The U.S. Department of Health and Human Services emphasizes inventories, transparent risk management, pre-deployment testing, independent review, monitoring and safe termination for high-impact AI. The World Health Organization’s principles call for human autonomy, safety, transparency, accountability, equity and continuous evaluation.
Governance should therefore be understood as an operating capability—not a policy document created after procurement. It must connect clinical leadership, patient safety, information security, privacy, compliance, legal, procurement, data science and frontline users.
Five priorities for executives
Figure 3. Five priorities for executives.
1. Build an enterprise AI inventory
Organizations cannot govern systems they do not know they have. The inventory should cover AI embedded in electronic health records, imaging platforms, medical devices, revenue-cycle tools, patient communications, workforce systems and vendor services. It should record intended use, accountable owner, developer, data inputs, affected populations and whether the output influences a consequential decision.
2. Match oversight to risk
An administrative drafting aid should not face the same controls as a system that influences treatment, benefits or financial assistance. A practical framework classifies systems by potential harm and establishes proportionate requirements for validation, approval, monitoring and human review. The goal is neither blanket restriction nor unchecked deployment; it is disciplined enablement.
Figure 4. Match oversight to risk: disciplined enablement.
3. Preserve meaningful human judgment
Human review must be more than a ceremonial click. Reviewers need authority to approve, modify or reject an output, access to the evidence required to evaluate it and training on the tool’s intended use and limitations. A governance program should also define when human review is mandatory, who is qualified to conduct it and how disagreement with an AI output is documented.
4. Create evidence that survives scrutiny
Healthcare organizations need reliable records of system versions, material changes, approvals, outputs, overrides, incidents and corrective actions. The federal HTI-1 framework likewise advances transparency for predictive decision support supplied through certified health IT, including source attributes and risk-management information intended to help users assess whether tools are fair, appropriate, valid, effective and safe.
5. Monitor performance and equity continuously
Pre-deployment testing is necessary but not sufficient. Changes in patients, workflows, interfaces or model versions can alter real-world performance. Monitoring should detect drift, subgroup disparities, unexpected failure patterns and uses outside the approved purpose, with clear thresholds for investigation, suspension or retirement.
A four-layer operating model
A robust governance architecture can be organized into four reinforcing layers:
- Verifiable records: Tamper-evident evidence of system versions, decisions, approvals, overrides and changes.
- Human oversight: Risk-based checkpoints that place qualified people in control of consequential decisions.
- Access and data controls: Role-based permissions, privacy safeguards and data-use boundaries suited to clinical environments.
- Continuous assurance: Ongoing monitoring for performance, misuse, security, drift and disparate effects.
Tamper-evident evidence of system versions, decisions, approvals, overrides and changes.
Risk-based checkpoints that place qualified people in control of consequential decisions.
Role-based permissions, privacy safeguards and data-use boundaries suited to clinical environments.
Ongoing monitoring for performance, misuse, security, drift and disparate effects.
Figure 5. A four-layer operating model.
This structure helps protect against more than regulatory noncompliance. It can expose unauthorized access, unsupported system changes, deceptive outputs, manipulation of records and use outside an approved purpose. It can also give boards, regulators, clinicians and patients a clearer answer to a basic question: Why should this system be trusted?
Governance enables adoption
Strong governance is sometimes presented as the brake on AI innovation. In healthcare, it is more useful to see it as the steering system. It allows leaders to distinguish promising uses from unsafe ones, expand successful tools with evidence and stop weak systems before harm spreads.
This does not mean every organization must wait for perfect regulation. Federal health-IT transparency rules, FDA oversight of AI-enabled medical devices, HHS risk-management priorities and state initiatives already point in the same direction: know the system, understand its purpose and limits, keep people accountable, monitor on-going performance and preserve evidence.
Figure 6. Existing rules and oversight already point in the same direction.
The organizations that build these capabilities now will be better positioned to adopt AI at scale. They will also be more prepared for a regulatory environment that is changing rapidly—as Colorado’s recent rewrite of its AI law demonstrates.
About the authors
Dr Ranse Howell is CEO of Regitech and founder of The Unified Sentient. His background spans law, alternative dispute resolution, social policy and the creative arts. He approaches AI governance with an emphasis on stakeholder alignment and on accountability systems that enable innovation rather than restrict it.
Laura Glennie is a healthcare services consultant. Laura has spent her clinical career focused on improving the quality of patient care and clinical excellence. Her knowledge of the Utilization Review process within large hospital systems, while maintaining a focus on the quality of patient-centered care, informs her thinking when anticipating the impact of AI and the future of healthcare.