If you’ve been anywhere near a data team, you already know the existential crisis happening right now. Here are just a few questions data leaders and our partners have shared with us:
- Why does data governance still feel like a slog?
- Can AI fix it, or is it making things worse?
- How do we move from governance as a roadblock to governance as an enabler?
These were the big questions tackled in this year’s Great Data Debate, where a powerhouse panel of data and AI leaders dove deep into dove deep into how governance needs to evolve.
Meet the Experts
This discussion brought together industry leaders with deep expertise in data governance, automation, and AI:
Tiankai Feng, Director of Data & AI Strategy at ThoughtWorks, advocates for human-centered governance and explores this philosophy in his book Humanizing Data Strategy.
Sunil Soares, founder and CEO of Your Data Connect, specializes in AI governance and regulatory compliance, navigating the challenges of large language models in modern data strategies.
Sonali Bhavsar, Global Data & Management Lead at Accenture, drives governance strategies for enterprise AI, emphasizing the importance of embedding governance from the start.
Bojan Ciric, Technology Fellow at Deloitte, focuses on automating governance in highly regulated industries, particularly financial services and AI-driven transformation.
Brian Ames, Head of Transformation & Enablement at General Motors, ensures data trust as GM evolves into an AI-powered, software-driven company.
The Three Biggest Data Governance Problems—And How to Fix Them
If there’s one thing that became clear, it’s that governance is at a crossroads. The old way—heavy documentation, rigid policies, and reactive fixes—simply doesn’t work in an AI-driven world. Organizations are struggling to keep up, and governance teams are often seen as roadblocks instead of enablers.
But why does governance keep failing? And more importantly, how do we fix it? The panelists zeroed in on three major problems — and the practical steps organizations need to take to get governance right.
1. Data Governance Is Always an Afterthought
“Governance usually only becomes important once it’s a little too late. Something has broken, the data is wrong, and suddenly everyone realizes, ‘Oh, we should have done governance.’” – Tiankai Feng
Let’s be honest: no one cares about governance until something breaks. It’s the thing that gets ignored—until a bad decision, compliance failure, or AI disaster forces leadership to pay attention.
This reactive approach is a losing game. When governance is treated as a last-minute fix, the damage is already done. The challenge, then, is shifting governance from an afterthought to an integral part of how organizations operate.
How to Make Governance Proactive, Not Reactive
- Make governance an enabler, not a clean-up crew. Instead of reacting to problems, governance should be built into processes from the start. Brian Ames explained how GM reframes governance as “consume with confidence” rather than imposing top-down rules. The goal? Making sure teams can trust the data they rely on.
- Start small and win early. Instead of rolling out governance across the entire organization, focus on a single, high-visibility issue where governance can deliver immediate value. As Tiankai put it, “Data governance takes time, but leadership expects instant results. You have to show impact quickly.”
- Tie governance to business outcomes. If governance is only about compliance, it will always be underfunded and deprioritized. Sunil Soares explained that successful governance programs are directly tied to revenue, risk reduction, or cost savings. If governance isn’t making or saving money, no one will care.
2. AI Is Exposing—and Amplifying—Bad Governance
“AI governance is exponentially harder than data governance. Not only do you need good data, but now you have to navigate regulations, explainability, and the risks of automation.” – Sunil Soares
The moment AI entered the chat, governance got even harder. AI models don’t just use data—they amplify its flaws. If your data is biased, incomplete, or lacks lineage, AI will magnify those issues, making unreliable decisions at scale.
AI governance isn’t just about ensuring quality data — it’s also about managing entirely new risks:
- Data bias: AI models make bad decisions when trained on bad data. If your data has blind spots, so will your AI.
- Lack of explainability: Many AI models act as “black boxes,” making it impossible to understand why they make certain predictions or recommendations.
- Automated chaos: AI agents are now making decisions autonomously, sometimes without human oversight. As Sunil warned, “The regulations are still talking about ‘human-in-the-loop,’ but AI agents are actively working to remove humans from the loop.”
How to Govern AI Before It Governs You
- Take a proactive approach to AI governance. Governance teams must anticipate risks rather than scramble to fix them after an AI-driven failure. This means aligning AI governance policies with existing regulatory frameworks and internal risk management strategies.
- Automate governance wherever possible. AI can actually help fix governance by auto-documenting metadata, lineage, and policies. “If governance is too manual, people won’t do it,” Bojan Ciric noted. “Automating metadata generation and anomaly detection saves time and makes governance sustainable.”
- Define AI guardrails before you need them. Organizations must create clear policies outlining what AI can and can’t do. This includes monitoring AI-driven decisions, enforcing retention policies, and ensuring AI outputs are accurate and explainable. Brian Ames described GM’s approach: “We need to define what our AI ‘voice’ can and cannot say. What’s its kindness metric? What are the things it must never do? Governance needs to ensure AI aligns with the company’s brand and values.”
3. No One Wants to “Do” Governance—So Make It Invisible
“If you lead with the word ‘governance,’ you’re going to run into resistance. The history of governance is that it’s painful, bureaucratic, and frustrating. We need to reframe it as something that enables people, not slows them down.” – Brian Ames
Nobody wants to be a data steward if it means spending half their time documenting rules in Excel. The biggest reason governance fails? It’s too manual, too slow, and too disconnected from the tools people actually use.
The reality is, governance can’t rely on manual processes. People don’t want to fill out spreadsheets or sit in governance forums that feel disconnected from their daily work.
How to Build Governance That Works, Without Anyone Noticing
- Make governance run in the background. Governance should happen automatically—things like lineage tracking, metadata collection, and policy enforcement should be built into workflows, not require extra effort.
- Bring governance to where people already work. Instead of making teams log into a separate governance platform, integrate governance into the tools they already use—Slack, BI platforms, engineering workflows. If governance isn’t embedded, it won’t get adopted.
- Use AI to take the burden off humans. AI can generate metadata, detect anomalies, and automate compliance tasks so people don’t have to. As Sunil put it, “People don’t want to do governance manually anymore—they expect AI to do it for them.”
Final Takeaways: How to Actually Make Governance Work
Governance is at a turning point. As AI reshapes how organizations use data, the old ways—manual, rigid, and siloed—won’t survive. The Great Data Debate 2025 made one thing clear: governance done right isn’t just necessary—it’s a competitive advantage.
The key to making it work?
- Embed governance into daily workflows. Governance can’t be a standalone process—it must be woven into the tools people already use, with automation handling compliance, lineage tracking, and policy enforcement in the background.
- Let AI govern AI. As AI adoption grows, it will take on a bigger role in monitoring policies, detecting violations, and ensuring transparency—reducing the burden on data teams while preventing AI from making unchecked, high-stakes decisions.
- Tie governance to measurable business impact. Instead of being seen as a cost, governance will be evaluated by its ability to protect revenue, improve efficiency, and ensure AI reliability. Organizations that prove governance delivers financial value will gain leadership support, while others struggle to secure buy-in.
- Invest in AI governance—now. Companies that delay will face mounting risks—regulatory, reputational, and operational. As Brian Ames put it, “AI governance isn’t optional—it’s the foundation for everything we do next.”
The future of governance isn’t just about compliance—it’s about scaling AI responsibly and unlocking data’s full potential.
Ready to build AI-ready governance?
Atlan makes governance seamless, automated, and built for the AI era. Book a demo today to see how Atlan can help your organization scale governance without the friction.
