The Future of Data Stewardship: A DGX Webinar Recap 

The traditional steward role is being replaced by AI. The new one is harder, more human, and most organisations haven’t named it yet. 

DGX recently hosted a panel on the future of data stewardship, bringing together two practitioners who are navigating the shift inside Australia’s largest banks. Luciano Vicente, with twelve years in data governance and currently at ANZ, and Mangala Wen, a former full-time data steward at NAB, joined DGX co-founder Chad Barendse to unpack what the role is becoming as AI moves into the centre of data operations. 

The conversation cut through the noise. Below is a recap of the key insights and the moments worth saving. 

The role being left behind 

Chad opened by drawing on his three years as a data steward at NAB — manual classification, chasing definitions across teams, and approval forums that produced decisions outdated by the time they were documented. The point: most stewardship models were built for a slower world. 

Luciano was direct about what comes next. 

“The traditional data steward, which became synonymous with policy access, manually approving definitions and chasing data quality issues after they happened, doesn’t scale and is quickly being replaced by AI.” 

Mangala put it sharper: 

“If stewards are still doing admin work in four or five years time, the organisation probably isn’t doing AI.” 

But she paired that with a real warning. AI is not a magic bullet. Lineage, metadata, and classification infrastructure still has to be built first. Without those foundations, AI on top doesn’t produce better governance. It produces faster garbage. 

“Sometimes we think AI is a magic bullet that will solve everything. Aspirationally, maybe. But first we need to make sure AI can actually help us do the basics.” 

The pattern Chad sees with clients reinforces this. Organisations buy a catalogue, plug in an LLM, and expect classification to “just happen”. Six months later they have the same backlog, only now it’s been auto-tagged inconsistently, and nobody knows which fields to trust. 

The new steward: judgment, context, and translation 

The strategic shift Mangala described was the most consequential thread of the conversation. The question moves from is this field accurate? to is this dataset good enough to support a lending decision or a regulator submission? That is a fundamentally different job. It requires judgment, context, and the ability to translate across audiences. 

“The best stewards I’ve learned from are the ones that can translate tech speak without pretending to be engineers, and speak the risk language without scaring people into compliance. You almost need to be the data whisperer to business and the business whisperer to data.” 

Luciano reinforced the point from a different angle: 

“Stewards having the context from the business, risk, technology and data enables them to ask the right questions.” 

Not the ones with the most tool access. The ones who hold context. 

The problems AI cannot solve: ownership and remediation 

This was the part of the panel that landed hardest. Mangala described the state of data issue remediation in most organisations as, in her word, aspirational. 

“The remediation of the data issues that I see is, the best word I can put it, aspirational.” 

AI can surface problems faster than ever. But deciding who owns the fix is still entirely human work. The RACI problem hasn’t gone away. 

“Where do data stewards fit on the RACI? Most of the time it would just be consulted. But somehow, data stewards are the ones being asked, ‘How come this data isn’t right?’ even though we’re never given the accountability or responsibility.” 

Accountability without authority. Luciano named what is arguably the deeper difficulty: 

“The hardest part of stewardship today isn’t tools or frameworks. It’s navigating organisational change and finding the right people.” 

And then refined it: 

“It’s not only the people who care, but who care AND can do something about it.” 

That distinction matters more than any platform decision. The stewardship programs that fail rarely fail because the framework was wrong. They fail because the people on the dotted line couldn’t actually move anything. 

Where AI governance fits 

Mangala was unambiguous about the stakes: 

“Data is an enabler of AI. Data governance really needs to sit at the table when we think about AI governance, because without the data, AI is effectively useless.” 

The steward’s emerging role here is convening — getting business, risk, technology, and security aligned before an AI tool is deployed, not after problems surface. Luciano added a counterpoint worth holding onto: 

“It’s dangerous to have a lot of decision makers in the same decision, because often the decision is compromised based on making everyone happy.” 

The goal isn’t just inclusion. It’s structured accountability. Too few voices and the organisation ships blind. Too many and it ships a compromise nobody owns. 

Where this lands 

The role evolves but it doesn’t disappear. The label might shift — data partner, data champion, capability partner. The function persists. 

The bigger question is who builds the governance brain for each business division: the context, the rules, the risks that AI will eventually need to act on, and who maintains it over time. Stewards are the obvious candidates. Most organisations haven’t decided that yet. 

Mangala’s own turning point was the clearest lesson from the panel. After months of no one acting on anything, she reversed her strategy entirely. She stopped framing stewardship as policy compliance and started anchoring everything to business outcomes. The organisations that figure this out will do the same. 

——— 

DGX is a data and AI governance consultancy based in Australia. We run regular webinars, training, and the annual DGX Summit in Melbourne — bringing together practitioners shaping the next generation of data and AI governance. To watch the full recording or learn about upcoming events, visit dgx.com.au.