AI governance in oncology is not a checkbox. It is a learning function.

The AI governance committee has become the beginning of the process

AI governance committees serve a critical evaluation process for healthcare organizations. AI and oncology are changing and advancing so quickly; the proper governance committee and structure are essential to evaluate the underlying architecture, development process, and ultimate success of the tool within clinical care. Technology, legal, compliance, ethics, clinical, data science, quality, patient safety, cybersecurity, procurement, and vendor-management leaders all have a role to play.

But the right committee is not simply a large committee. Its membership should reflect who the technology will affect. For an oncology or patient-facing tool, that means relevant clinicians, patients, caregivers, and people from underrepresented populations should all have a voice.

The committee also needs a clear charge: What evidence is required? What does success look like? How will the tool be maintained? Who has the authority to pause or retire it?

AI governance has to continue throughout the life of the model. Health systems need to understand performance across populations, disease states, and care settings. They need to monitor whether the tool changes clinical behavior and whether those changes are beneficial, neutral, or harmful. They need to track new evidence and understand whether the original decision still makes sense.

From reviewing tools to learning from them

We often treat governance as a review function connected to procurement or contracting. The more important question is whether the tool solves a meaningful problem—and whether the organization can demonstrate its value over time.

That requires a stronger informational foundation. Performance data needs to be trusted. Measures need to be relevant to the people using the tool. And results need to arrive while decision-makers can still act on them.

A vendor’s technical performance measure may not be the measure that matters most to a health system. A model may perform well in development but be less useful for a particular population or workflow. Governance committees need the authority and data access to see those differences.

Different tools require different levels of scrutiny

Another challenge is treating every AI tool as if it carries the same level of risk. It does not.

A tool that influences diagnosis, treatment, or patient safety may need intensive oversight. An operational tool with limited impact on patient care may need a lighter process. Applying the same level of friction to both can slow responsible innovation and dilute attention from the tools that matter most.

Risk-proportionate governance gives health systems a way to move quickly while preserving accountability. It helps committees focus on the questions that matter most:

Who could be affected? What could go wrong? What evidence would change our decision? What signals will tell us that the tool is helping—or that we need to revisit it?

Less approval. More accountability.

The future of healthcare AI governance may require fewer one-time approvals and more ongoing accountability.

Health systems need to connect learning from individual tools rather than evaluating every system in isolation.

In oncology, this connected view is especially important. At CancerLinQ, we think about building a connected learning network around cancer care—bringing together clinical data, treatment patterns, patient outcomes, and real-world signals so organizations can improve continuously.

The purpose is not to replace clinical judgment. It is to place better evidence in front of clinicians when they are making decisions for individual patients.

AI governance is entering a different stage of maturity. The objective is no longer simply to establish a committee or approve a model. It is to create an operating system for learning—one that can monitor performance, recognize change, share knowledge, and act before problems become permanent.