Why Proactive Governance Is A 'Continuous Pipeline' for Enterprise AI

AI is compelling financial institutions to rethink data governance. For decades, the rules were clear, built on pillars like data lineage, cataloging, and quality. But the introduction of AI is challenging that established playbook. Now, it's even raising new concerns about model reliability and ethical outcomes.

The transformation is one that Ashish Dibouliya (PhD), a Senior Enterprise Data Architect and Managing Director at Webster Bank, knows well. With over 19 years of experience in enterprise data warehousing for the financial services industry, Dr. Dibouliya has spent most of his career leading large-scale data projects at companies like GE Capital and ADP. Today, he's a certified Six Sigma Black Belt with domain expertise in consumer and commercial banking, risk management, and federal reporting. According to Dibouliya, the industry is already in the midst of rewriting its own governance rules.

"Financial institutions are redefining governance, looking beyond those four pillars and bringing the ethical and accuracy parts into the mix," Dibouliya said. "In the past, the accuracy of our predefined tools was well-accounted for. Now, that accuracy is becoming questionable." But the solution isn’t faster governance, he continued. Instead, it's building a disciplined, proactive quality assurance process that happens long before any AI model touches a live system.

• Speed through safety: With the help of a dedicated governance team, the approach vets every model to verify its reliability from the start. "The most important part is adopting the right and reliable AI models," Dibouliya said. "Once we ensure the reliability of these models, it does not really impact the speed. Once that part has been established, it's more or less a matter of tuning or grounding the model."

Get that right, and the subsequent architectural challenge becomes far less daunting, Dibouliya continued. From there, the task becomes a simple move in tooling.

• Everything and nothing changes: The shift from traditional providers to AI-native solutions modernizes the methods of governance without altering its core principles, Dibouliya explained. "We are not changing anything architecturally. We are shifting towards AI rather than traditional tools for tasks like data lineage and cataloging. What is important here is that when I get data lineage or data cataloging using AI, the model selection and grounding of the AI model are crucial. I think everything else remains the same."

In Dibouliya’s view, this operational evolution also requires governance itself to assume a new role. "It should be more proactive and work as a parallel entity rather than just a gatekeeper. Now, that is happening in this industry. It is becoming more prominent where governance works as a continuous or parallel pipeline, not just a gatekeeper."

• The human backstop: For Dibouliya, this change makes human oversight non-negotiable. "AI can accelerate some human work, but it needs continuous human monitoring or intervention. When it comes to governance and explaining this to auditors, you still need human support to determine whether the adopted AI tools are efficient or not. Human intervention or monitoring is very crucial, at least for now."

Ultimately, Dibouliya’s vision of the future is a pragmatic one. Here, he offered the familiar example of Anti-Money Laundering (AML), where specific events trigger an alert that requires human investigation. After sharing a personal story about having his own card blocked while traveling, he explained why false positives are still a common occurrence. With AI, however, the same event can be enriched with additional context—like a customer's known travel patterns—enabling the system to make smarter decisions, reduce reaction time, and improve accuracy.

While the system's foundational backbone remains intact, it's now supported and refined by an intelligent AI layer. "The role AI is going to play is providing more relevant or refined information," Dibouliya concluded. "AI will provide us with the right insights, and reaction times will be reduced. Architecturally, everything remains the same. You're just going to add AI as an additional supporting element to the whole architecture."