What the Titanic’s Stewards Understood

The RMS Titanic carried an unusually large stewarding staff for its time. Of the roughly 885 crew aboard, around 320 were stewards, one of the largest groups on the ship. When the ship struck the iceberg, these stewards were critical in saving many lives, waking sleeping passengers and telling them to dress warmly. They carried children, calmed the frightened, and guided the lost through a maze of corridors. Without this assistance, many people would have remained below decks far too long. Most of the male stewards died doing it. Many were last seen helping launch the boats they would never board.

The officers could issue orders, but execution fell to the stewards, cabin by cabin, person by person. They became the human bridge between policy and action.

What We Asked of the Data Steward

The Titanic reminds us of what stewardship truly means. A steward is someone entrusted to care for something they do not own, standing closest to the people who rely on them and serving with quiet responsibility. There is something deeply human and noble in that idea, and we were right to want it in our work.

When Data Governance appeared on the horizon, stewardship came along with it. So what did we ask of the steward in the world of data management and Data Governance? Someone entrusted to care for the data, protect its quality, see that it is used properly, and help others find and understand it. It was about custodianship, care, and accountability on behalf of others.

In almost every place I have worked, stewards were handed the responsibility without the training to meet it. At one company I ran a survey on stewardship, and more than 50% said, “I wasn’t aware I had that responsibility.” Others only gave it time when someone escalated an issue. The rest could not describe what stewardship asked of them. A 2023 Gartner survey pointed the same way: fewer than half of data and analytics leaders, 44%, said their teams effectively deliver value to the business.

A Long-Standing Question in Data Governance

Why does data stewardship rarely move the business? It is a question many Data Governance leaders avoid, because it questions an assumption the whole practice rests on. Stewardship is not failing. Most organizations are simply building stewardship activities, not stewardship outcomes.

That is what stewardship keeps becoming, administrative work. Stewards spend their days defining terms, updating catalogs, and closing compliance tasks. We count stewards assigned, terms defined, meetings held, and call it progress. All of it measures activity.

The business leaders above are asking about something else entirely. Fewer customer complaints, faster product launches, cleaner regulatory compliance, and above all, higher trust in the data analytics. The governance office is winning at governance while the business still waits on results.

A few things keep it that way. Stewardship becomes a part-time duty, squeezed beside the day job, its benefits too far downstream to feel. Organizations invest heavily in the toolstack, yet far less in the people meant to use it. So stewardship solves data problems when the business needs business problems solved.

This is the shift I keep coming back to. Stewardship succeeds when we treat it as a business performance capability, judged by the business it improves.

What AI Hands Back to the Steward

The shift is already here. Data platforms are now shipping autonomous AI agents that document assets, recommend classifications, and keep the glossary current on their own. Gartner expects 40% of enterprise applications to carry task-specific agents by the end of the year. The administrative weight that came to define the steward’s day is being lifted, quietly and quickly.

I find this hopeful, for a reason most people miss. The fear is that AI replaces the steward. I see the opposite. AI is taking the part of stewardship that was always a burden, the tagging, the form-filling, the chasing of blank fields, and leaving behind the part that always mattered. Judgment, context, and knowing whether the data can be trusted, then standing behind that answer.

The machine can document the dataset. It cannot care about it. That distinction is the whole future of the role.

The Collaboration Between AI Agent and Data Steward

Here is what worries me. We have a long habit of funding the tools and leaving the people behind, and AI hands us the perfect reason to do it again. The AI agent governs now, the thinking goes, so the steward matters less. I believe that gets it backward.

So here is the discipline I would hold us to. We measure the agentic AI steward by the business outcomes it moves, by whether trust, risk, and decisions actually improve. This is where the AI agent earns its place twice over. Outcome metrics were always the truer measure, yet tracking them was so laborious that most teams settled for counting activity instead. The AI agent makes that hard measurement possible, gathering the evidence and watching over time whether the business is better off.

This is collaboration. Each side does what it does best. The AI agent carries the administrative load. The human data steward stays the decision maker, reading what the AI agent surfaces, judging what it means for the business, and standing behind the call. The AI agent supplies the speed and the reach. What the steward adds is judgment, and the accountability that comes with owning a call. Neither delivers the business outcome alone.

The Future of Stewardship in the AI Realm

AI raises the data steward. That is the future of the role, and it lifts them higher than the role has ever stood. As AI agents take on the repetitive work, tagging assets, documenting metadata, monitoring activity, measuring outcomes, the steward steps into a different place in the organization, now working under the business wing. That frees them for the judgment, the context, the decision that drives a business outcome.

We spent years asking data stewards to look after the data. The role was always larger: to be the reason a business can trust the data behind its products and its AI models, and decide with confidence. Agentic AI can finally make that real.

Let the machine carry the work, and the steward carry the trust.