Data Governance for AI Is Under Pressure: What the Data — and Experts — Are Telling Us

Key Takeaways

• Without a formal data governance program, trust in organizational data is essentially a coin toss. That gap has real consequences as AI systems scale decisions at speed. 

• A data governance program and a data strategy must work together; governance built in isolation from business objectives stalls quickly, and organizations that align both see significantly stronger AI outcomes. 

• AI governance isn’t a separate discipline, but an extension of existing data governance. Organizations getting it right are embedding governance into business processes, not treating it as an add-on. 

For data and analytics leaders today, AI has raised the bar for everything underneath: data quality, data integration, and above all, data governance. Not as a compliance checkbox, but as a genuine strategic foundation. And the organizations that haven’t treated it that way are starting to feel that pressure sharply. 

That was the central premise of our recent Data Integrity & AI Forum session, AI Is Stress Testing Data Governance: Are You Seeing the Same? — a conversation I had with three people who bring very different, and very complementary, perspectives on this challenge: Nicola Askham, the Data Governance Coach; Kevin Guyot, VP of Data Analytics at Lake Michigan Credit Union (LMCU); and David Woods, SVP of Global Services at Precisely.  

I want to highlight a few of the themes that stood out most, and the questions I think every data leader should be sitting with right now. 

Why Data Governance Produces Trust, and What Happens Without It 

One of the clearest signals in this year’s 2026 State of Data Integrity and AI Readiness report — published by Precisely in partnership with Drexel University’s LeBow College of Business and based on a survey of more than 500 data and analytics leaders — is the relationship between governance programs and data trust.  

 

Data leaders at organizations with formal governance programs report “high/very high” trust in their data at a rate of 71%, compared to 50% at organizations without a governance program in place. In other words, without a governance program, trust in data is essentially a coin toss.

That has direct consequences when AI enters the picture. Models and agents that consume untrustworthy data produce flawed outputs at scale, with speed. 

Kevin’s experience at LMCU gets to the heart of why that gap exists. For him, the difference between organizations that trust their data and those that don’t isn’t about documentation, but whether governance is embedded in how decisions actually get made: 

“It’s not just something that sits on a page, but it’s got true ownership to it, and it’s embedded in their day-to-day decision-making. And so when we see that and when we actually do that, we start to really trust the data.” 

David adds that the driver for governance has shifted significantly over the past two to three years. What used to be an industry-specific concern, concentrated in financial services, healthcare, and heavily regulated environments, has become a universal requirement: 

“What we’re seeing now is trust has really become the universal requirement for governance, and AI is only accelerating that. Because as you know, we feed a model bad or uncontextualized data, the outputs erode trust really fast. Governance has really gone from a compliance thing to something every business needs if they want their data to actually work for them.” 

The survey’s confidence-reality gap reinforces this urgency. While 87% of data leaders describe their organizations as AI-ready, 43% cite data readiness as a top barrier to AI success.  

That is not an inconsistency. It is a measurement problem. Organizations are evaluating readiness against outdated criteria, and Agentic AI, where autonomous systems make decisions without direct human oversight, is making those gaps immediately consequential. 

Does Your Governance Program Have a Data Strategy Behind It? 

Of the organizations surveyed, 83% say they have some form of a data governance program in place. But with data readiness cited as a key AI barrier, the question isn’t whether governance exists, but whether that governance is doing the right work. 

Our live audience poll flagged a consistent theme: lack of ownership and accountability. David pointed to something he sees repeatedly when organizations ask for help: data governance built without a clear connection to data strategy.  

“Organizations that tend to treat them as separate workstreams really can’t articulate the value and point to outcomes where governance was an accelerator and an enabler for the successes,” David said. “The moment those two things connect, that’s where you see the trust scores and the business results that follow.”

It’s also crucial to connect governance to business outcomes. Nicola put it bluntly: it doesn’t matter how good your governance program looks on paper if nobody in the business can explain what problems it solves.  

She recounted a recent coaching session where a client listed “better understanding of our data” as a benefit of governance. Nicola’s response was simply: “So what?”  She explained, “You will, but that’s not actually what your business people want to hear. What will help them if they understand the data? So, making it back to the business alignment: What business problems will you solve? What business value will you deliver? – it’s really important.” Kevin described what top-down alignment looks like in practice at LMCU, and it doesn’t start with governance. It starts with business goals, business priorities, and business objectives, with governance as the mechanism that makes those outcomes achievable: 

“Ultimately it’s telling them the why. When this data is clean, what is that going to allow you to do? What kind of decisions can you actually make faster? How much time do you spend looking at a report and then another report trying to understand why these numbers don’t match and spend more time analyzing data than making decisions on data?“ 

The four-profile framework from the research illustrates the compounding effect of alignment. 

• Innovators (data strategy and data governance program) | 72% high data trust 

• Experimenters (data strategy without data governance program) | 61% high data trust 

• Planners (data governance program without data strategy) | 40% high data trust 

• Laggards (neither) | 0% high data trust 

To reach the Innovator designation, organizations must intentionally connect governance to strategy as parallel workstreams rather than sequential ones. 

How Do You Fit AI Governance into an Existing Data Governance Program? 

This is the most hotly debated question I hear from data leaders right now, and the research reflects that tension.  

• 40% of organizations are extending existing data governance to include AI governance 

• 23% are building a separate framework.  

My honest view is that it’s likely not a pure either/or, and the panel’s experience bears that out. 

Nicola has seen data governance teams increasingly asked to absorb AI governance responsibilities, often without additional resources or the skills that AI governance specifically requires: “The data governance people feel comfortable they could do the [data quality] part of AI governance, but they feel totally at a loss about how they do the model validation, how do you check for bias — because this isn’t their skill set. A lot of them feel like they’d rather it was done by them than not at all … but they’re feeling rather overwhelmed by it.”

Kevin’s approach at LMCU is to consolidate governance councils wherever possible, and to ask a specific set of questions before any AI use case moves forward. Those questions go beyond technical feasibility: “Does it even make sense to use AI in this capacity? If we decide it does, do we even have the data of sufficient quality to allow AI to actually act in that way? And ultimately … just because we can, should we?” 

He emphasized the need to ensure alignment to your organization’s core values, ethics, and principles through it all. 

David’s perspective is grounded in what he’s seen across hundreds of customer conversations: most organizations started AI governance with a list of approved tools and a use case evaluation process, and stopped there. The data piece came later, often too late: 

“If your existing governance framework has context, lineage, quality rules — that’s exactly what AI needs. Start there and pull that in, but the difference is it has to evolve. Kevin said the unique piece we’re adding with AI governance is the “Should we?” And that includes new considerations and new personas, new roles around ethics, legal, HR that typically haven’t been part of traditional [governance] councils.” 

The broader shift David sees, which is validated by the research, is that governance has to move from being a separate governance team activity to something embedded in business operations. Nine out of ten AI use cases still fail to reach production, David noted, and the root cause is almost always the data. Organizations that have gotten use cases into production are the ones who learned early that governance is not a gate on the way to AI deployment, but the foundation underneath it. 

Where to Start with Data Governance for AI 

As we wrapped up the session, I asked each panelist what they would tell any data leader in the room to prioritize over the next twelve months. The themes were consistent. 

Kevin pointed back to governance as the starting point, because it’s the mechanism that makes everything else possible: 

“When you do that, it creates that foundation that drives enough business context for these AI models to be able to consume, to be able to make reliable choices and not hallucinate. Not that they never will, but if you can’t even get to common language, common terminology, you can’t trust the quality of your data — you’re not going to be able to do that unless you drive that with a data governance program. You’re never going to be ready for AI.”

Nicola pressed the point that investment has to match priority, and right now, in too many organizations, it does not. She’s seen data governance leaders made redundant as part of AI investment programs, which misses a fundamental truth: the two aren’t competing priorities, but dependencies. 

David’s recommendation for the second half of 2026 was direct: improve your business semantic layer: the contextual layer that gives AI systems a coherent, governed view of what the organization’s data means. It is the mechanism through which governance becomes intelligence: 

“If you can have a contextualized semantic layer that provides and exposes that data via our governance program with context, you will leapfrog your competition, and your data program will be a key pillar underpinning anything that the organization wants to do. For me, six days a week, twice on Sunday, I will be focused on improving and ensuring my semantic layer is not only accessible by business and IT users, but importantly accessible by Agentic agents.” 

The Organizations That Get Data Governance Right Will Successfully Scale AI.  

What came through clearly in both the report and our conversation is that data governance is no longer a program you run in parallel to your AI strategy. It’s the thing your AI strategy runs on.  

The organizations making real progress are building governance and strategy together, embedding it into how work actually happens, and asking the harder questions — not just whether AI can be deployed, but whether it should be, and whether the data behind it is trustworthy enough to act on. 

The disconnect between organizations that do this well and those that don’t comes down to the fundamentals. And the data makes clear that the strongest foundations start with governance. 

Read the 2026 State of Data Integrity and AI Readiness report to benchmark your own organization against more than 500 data and analytics leaders. And watch the full panel discussion, The Data Integrity & AI Forum: AI Is Stress Testing Data Governance: Are You Seeing the Same?, to hear Nicola, Kevin, and David go deeper on these themes.