Making Data Governance Stick: How Healthcare Organizations Sustain AI Success

By Mark Leifer, Data + Analytics Manager, Tegria

Healthcare organizations are making real progress with AI. Pilots are launching. New tools are rolling out, including capabilities embedded directly in EHR platforms. There is real momentum. 

In a recent HIStalk Readers Write article, I wrote about why data governance and executive sponsorship are foundational to realizing value from AI. That conversation resonated with organizations trying to move beyond experimentation and into something sustainable. 

But there is a second challenge that comes up just as often. 

A lot of healthcare organizations can get governance started. They can define roles, stand up a council, create some policies, and hold a few meetings. The harder part is keeping it going once the initial push wears off. 

That is where many efforts start to lose altitude. 

Meetings get canceled. Priorities shift. Stewardship becomes a side job. Governance starts to feel like a project that had its moment instead of a capability the organization relies on every day. 

And when governance fades, AI value usually fades with it. 

Why Governance Efforts Lose Momentum 

Most governance efforts don’t lose traction because the original idea was bad. They stall because they were treated like a one-time project. 

Teams do the setup work. They assign stewards. They publish a charter. They stand up a committee. Then attention moves to the next strategic priority, and governance slowly drifts into the background. 

A few warning signs show up again and again:

• Executive attention moves elsewhere

• Teams cannot see a clear line between governance and outcomes

• Stewardship is assigned, but not truly supported

• Metrics are either too vague or nonexistent

• Governance lives in documentation, not in day-to-day work

If that sounds familiar, you are not alone. This is a common stage for organizations that are still maturing their approach to data and AI governance.

Shift the Mindset: Governance Is Not a Project

If healthcare organizations want governance to last, they have to stop treating it like something that gets “implemented” and then checked off.

Data governance is an operating discipline. 

Healthcare data is always changing. New use cases emerge. New data sources are introduced. Workflows evolve. Regulations shift. AI models and vendor tools need oversight after go-live, not just before launch. 

That means governance has to be built into how the organization works. It cannot sit on the side as a separate effort that people remember only when there is a data issue, a compliance concern, or a model that starts producing questionable results. 

Anchor Governance in Business Value 

One of the fastest ways to lose momentum is to talk about governance only in governance terms. 

Most leaders are not asking for “better governance.” They are asking for better outcomes. They want better patient care, less rework, reduced risk, stronger compliance, more confidence in reporting, and a clearer path to using AI responsibly.

Governance has to connect directly to those goals.

In healthcare, that connection is not hard to make:

• Cleaner data reduces rework in reporting and analytics

• Clear ownership speeds up issue resolution;

• Standard definitions improve trust in dashboards and KPIs

• Better stewardship supports safer AI deployment;

• Strong controls reduce privacy, compliance, and operational risk

When governance is framed as an enabler of outcomes rather than an administrative layer, it is much easier to sustain attention and investment.

Keep Executive Sponsorship Active

Executive sponsorship matters at the beginning, but it matters just as much six months later.

Governance loses momentum when leaders approve it once and then disappear. It stays relevant when leaders continue to show that it matters. 

That does not mean executives need to manage the day-to-day work. It means they need to stay visibly connected to it. 

In practice, that can look like: 

• Reviewing governance KPIs in leadership meetings  

• Participating in governance councils at key intervals  

• Reinforcing governance expectations in strategic discussions  

• Backing stewardship decisions when priorities conflict  

• Asking whether AI tools have gone through the right review and monitoring steps  

When governance shows up in executive conversations, the organization pays attention. It signals that governance is not optional and not temporary. 

Build Governance Into Everyday Workflows 

Governance does not stick when it lives only in charters, slide decks, and policy documents. 

It sticks when it becomes part of the normal flow of work. 

That means asking where governance should show up in existing processes, not where a separate governance process can be added on top. 

For healthcare organizations, that might include: 

• Governance review as part of AI intake and approval  

• Data quality checks embedded into reporting and analytics workflows  

• Defined ownership for key data elements and business terms  

• Clear escalation paths for data issues  

• Steward participation in projects that introduce new data, definitions, or models  

• Periodic review of live AI solutions to confirm they are still performing as intended and being used for the approved purpose  

This is one of the biggest differences between governance that fades and governance that lasts. The durable programs are not built around extra meetings alone. They are built into the processes people already rely on. 

Empower and Support Data Stewards 

Stewardship is one of the most important parts of making governance sustainable, and one of the most mishandled. 

Too often, organizations name data stewards but do not define what that means, how much time it requires, or what support they will receive. The title exists, but the operating model does not. 

That is a recipe for burnout and inconsistency. 

To make stewardship sustainable: 

• Clarify expectations and scope  

• Define decision rights and escalation paths  

• Provide training, tools, and practical guidance  

• Recognize contributions and outcomes  

• Connect stewardship work to business priorities people care about  

When stewardship is embedded into existing roles and workflows, governance becomes much more durable. It starts to feel like part of how the organization runs instead of a separate program asking for favors. 

Add Post-Deployment Monitoring to the Governance Model

This is the biggest addition I would make to the conversation now. 

A lot of governance models still focus heavily on the front end: intake, approval, risk review, and deployment. That matters, but it is not enough anymore. 

For AI in healthcare, go-live is not the finish line. 

Organizations need a practical way to monitor AI after deployment. That includes asking whether the solution is still being used for its intended purpose, whether the underlying data or workflow has changed, whether it is still delivering the expected value, and whether any new concerns have emerged around bias, safety, compliance, or trust. 

This does not have to be overly complex. In many cases, a lightweight recurring review can go a long way. 

A simple post-live monitoring approach might include: 

• A named business owner for each live AI use case  

• A regular attestation that the use case, workflow, and source data have not materially changed  

• A small set of performance, adoption, and risk indicators  

• A trigger for reevaluation if the model, workflow, or regulatory context changes  

• A mechanism for users to raise concerns when outputs do not look right  

That is how governance stays connected to reality. It is also how organizations avoid the trap of assuming that an approved AI solution will remain safe, useful, and trusted indefinitely. 

Measure What Matters 

Governance programs need proof of value if they are going to survive shifts in priorities and leadership attention. 

That does not mean building a giant dashboard with 40 metrics—that usually makes things worse. 

A smaller, more practical set of measures is better. Especially early on. 

For example: 

• Percentage of critical data assets with defined ownership  

• Participation and attendance in governance or stewardship forums  

• Number of data issues identified and resolved  

• Data quality improvement in key domains  

• Percentage of AI use cases reviewed through governance  

• Percentage of live AI solutions with an active monitoring cadence  

• Training completion for stewards and governance participants  

The goal is not measurement for measurement’s sake. It is to show that governance is active, improving operations, and reducing risk. 

Good metrics also reinforce accountability. They help leadership stay engaged and help teams see progress over time. 

Use Stories, Not Just Scorecards

Culture plays a major role in whether governance lasts. 

People engage more when governance feels tangible. Metrics help, but stories often land better. 

Share examples like these: 

• A data issue was caught before it affected a clinical or operational decision  

• A report became more trusted because key definitions were standardized  

• A governance review improved an AI use case before broader rollout  

• Duplicate or incomplete records were reduced because stewardship was clearly assigned  

• A live AI tool was reevaluated after workflow changes and adjusted before problems spread  

Stories make governance real. They help people see that governance is not about slowing work down. It is about helping the organization move with more confidence. 

What Lasting Governance Looks Like

The healthcare organizations that get long-term value from AI are usually not the ones with the most tools. They are the ones that make governance part of how they operate. 

They align governance to business priorities. They keep leadership engaged. They formalize stewardship without overcomplicating it. They build governance into workflows. They measure progress. And they monitor AI after deployment instead of assuming the hard part is over. 

That is what makes governance stick. 

Not perfection. Not bureaucracy. Not a bigger committee. 

Consistency. 

Final Thought 

AI will continue to expand across healthcare. The organizations that benefit most will not be the ones that move fastest without guardrails. They will be the ones that build governance into the way they make decisions, manage data, and support change over time. 

Getting governance started matters. 

Making it last is what turns governance into value. 

Take the next step. Talk to our experts about assessing and strengthening your data governance strategy for AI.