Building a solid data foundation for AI applications is an essential starting point that sets the stage for additional AI governance efforts in organizations.
During a webinar on the future of AI governance last week, Lisa Wintrick, an executive advisor at consulting firm First San Francisco Partners, said expectations that the AI era would be “data’s moment” are playing out in deployments at the organizations she works with. “The better your data is, the better everything else is,” she said.
“The better your data is, the better everything else is.”
Lisa Wintrick
Executive advisor, First San Francisco Partners
Wintrick said data teams in those organizations are focusing heavily on the basics of data governance — for example, tracking the lineage and validating the quality of the data used by AI tools. But she noted that interest in master data management (MDM) has also increased among her clients in recent months.
While MDM naturally dovetails with data governance, its complexity relative to the expected business benefits limited its adoption in the past, mostly to large enterprises. Now, comprehensive metadata and master data are both key components of an AI-ready data ecosystem that provides AI with required context and appropriate governance controls, First San Francisco founder and CEO Kelle O’Neal said in the webinar.
O’Neal and Wintrick said creating that ecosystem is critical to AI success — but it isn’t the endpoint in developing a mature AI governance strategy. For example, they said governance leaders should also do the following:
- Connect data and AI governance decision-making. In many organizations, “we’re still seeing AI governance over here and data governance over here,” Wintrick said. “Bringing those two together will make sure that everyone is going in the same direction and that your AI initiatives and your data initiatives are supportive of each other.”
- Better align governance policies and practices. Pointing to survey data published by EY US in September 2026 showing that 47% of organizations had bypassed their AI governance process for urgent deployments, O’Neal said doing so turns a governance policy into “merely a guideline.” Wintrick acknowledged that governance leaders might need to go along “when the batphone rings” and a senior executive asks for a fast-path deployment. But they should create requirements for such requests and weigh the risks vs. the potential business rewards, she said.
- Adopt what they called a decision stewardship process. Similar to how data stewardship maintains data assets and enforces related governance policies, O’Neal said decision stewardship — a nascent concept First San Francisco is promoting — would examine AI outputs for accuracy and policy compliance. Assuming that will happen organically “because it’s the right thing to do is not a good play,” she added.
Governing beyond the data layer
In a report on the state of data and AI governance published in September, BARC analysts Kevin Petrie and Timm Grosser wrote that initially focusing on the data layer is natural for governance teams, given their familiarity with it and data's pivotal role in successful AI. That’s reflected in the BARC survey of data and AI leaders underlying the report: 62% of the 234 respondents said their organization has adopted AI governance controls on data, vs. 45% and 43% at the AI model and infrastructure layers, respectively.
Moving governance processes deeper into the architecture that underpins AI systems is a must, Petrie and Grosser said. That includes addressing unstructured data, which provides critical business context for AI models and agents but has been ignored by many data governance teams until now. In addition, they said teams should extend data quality controls to items such as metadata, embeddings, retrieval pipelines and internal knowledge bases.
Craig Stedman is a senior reporter on the TechTarget Data news team, covering analytics, data management and data backup topics. He has more than 40 years of experience as an enterprise IT writer and editor.
