For twenty years I did what I was told.
Go stand up a governance program. Go build the master data. Make the bosses happy. I was good at it, and I rarely stopped to ask whether the thing I was building would actually get used.
Then I built master data management for a manufacturing plant, and nobody came.
I can tell you exactly why. It wasn't the tool. It wasn't the model. It was me. I built the capability and skipped the hard part: the change management, the rollout, the work of getting people to actually change how they worked. I shipped a system and assumed the value would speak for itself. It didn't. It sat there. That failure taught me more than any success ever did, and it is the reason I can say what comes next without flinching.
Here is what I believe now.
The era of governance theater is ending.
And governance theater is exactly what it sounds like. It's all the work that looks like governance and changes nothing. The council that meets and decides nothing. The framework that impresses an auditor and helps no one. The catalog with four thousand entries nobody opens. The policy sitting on a shared drive nobody follows. Lots of motion, no movement. We built programs to look responsible instead of to help people decide, and everyone in the field knows it even if they won't say it out loud. A catalog nobody opens isn't governance. It's decoration.
AI agents are about to expose all of it. That isn't a threat. It's the best thing to happen to this field in twenty years.
Because it finally forces the question we have spent two decades avoiding: does this work deliver value, or doesn't it? AI is going to do overnight the work that used to take us months. The classification, the tagging, the lineage, the definitions we argued about in rooms for weeks. When that happens, the programs built to look busy get shown for what they are, and the work that actually moves a decision finally gets its turn.
Governance moves from the back office to the front office. That is the whole shift.
And the data steward moves with it. If you are a steward reading this, hear me clearly. Your job is not going away. It is getting promoted. The person who spent years writing rules nobody read becomes the person who directs the agents everybody depends on. You declare the outcome. You set the rules the agent runs against. You validate the result. That is a bigger job, not a smaller one. But it comes with one condition: you have to build the skills, and you have to understand how AI actually works. Do that, and you don't just survive this shift. You lead it.
Because the method itself has to change. Data governance cannot take years anymore. It has to move at the pace of AI, and that means weeks, not years. We need active catalogs, not passive ones. Assets that do something, not documents that sit. And we don't get there by putting more people in more meetings. We get there by letting AI suggest the definition, suggest the tag, suggest the data quality rule, and having a human validate it, instead of sitting in a room for months trying to author every rule by hand. Stop authoring. Start validating. Let the machine draft, and let your people judge. Do more with less, or get exposed by the people who do.
But be clear about one thing, because this is where most AI projects are going to fail. Agents are not magic. An agent is only as good as the rules you give it and the data you point it at. Give an agent no definition of a customer and it will make one up. Give ten agents no definition and you get ten different answers, all confident, all wrong in their own way. So here is what an agent actually needs to succeed, in plain terms. A clear job, which is the outcome you want. Clear rules, so it knows what good looks like. Clean data underneath it, so it isn't building on sand. A memory, so it doesn't start from zero every time you ask. And a name on the work, so a human owns the result. Give it those five things and it flies. Skip them and all you have done is automate the mess faster.
Now the part I am angry about.
None of this works without the executives, and too often the executives are the problem. They stand on stage and call data a strategic asset, and most of them cannot tell you what that sentence actually means. They fund the program at a fraction of what it needs and expect miracles. They will not mandate it from the top, and then they blame the data team when nobody adopts it. Let me be plain about something I learned the hard way at that plant. You cannot build data governance from the bottom up. It has to come from the top, with real support and real air cover, and it takes time to do right. A slogan is not support. A budget line that gets cut first is not support.
And to the executives who are resisting the faster way even now: I don't get it. I am telling you there is a way to do this in weeks instead of years, and you are fighting it. This is not about buying a technology. It is a mindset shift. How can you do more with less? How can you let the machine draft and let your people judge? That is the question, and the companies that answer it in 2026 are the ones whose AI bets actually pay off. Everyone else is buying catalogs nobody opens and calling it strategy.
I did not always talk like this.
For most of my career I kept my head down and delivered what I was asked to deliver. One day cracked that open.
I was standing in a boardroom in front of the CEO and the rest of the C-suite. I could not tell you today what the question even was. What I remember is that it was aimed at two of us, the CIO and me, and the CIO answered it with something that was flat out untrue. I had been taught my whole career that you present a united front in front of leadership, so that is what I did. I did not lie. But I did not correct him either. I let his version stand.
The moment the meeting ended, I went straight to my boss and told him the CIO had lied. What he said back has stayed with me ever since. When you have a platform, you speak your truth. What you genuinely believe and what you actually know, even when it means correcting the person standing right next to you. Staying quiet in that room, he told me, was its own kind of lie. And he was right.
That became a guiding principle for me, and it is a large part of why this manifesto exists at all.
Two other things pushed me the rest of the way. Personally, I finally accepted that I have ideas worth saying out loud, and that I can disagree without being disagreeable. Professionally, I earned a data science degree I had been sitting on, and I watched AI turn the thing I have always been most curious about, the place where data science meets data governance, into the most important problem in the enterprise. I could fight that or I could embrace it and enjoy the ride. I am choosing the ride. This is foundational, and it is going to change data governance from the thing we know today into something far better.
So here is where I stand.
I am staking my career on this. Not as a hedge. As a conviction. I believe the era of governance theater is ending. I believe the steward becomes the orchestrator. I believe the companies and the people who see it first are the ones who win. I might be early. I will be wrong about some of the details. I am betting on the direction anyway, because I have spent twenty years inside this field and I can feel the ground moving under it.
For twenty years I did what I was told. Not anymore.
If your job is about to be rewritten, don't wait for someone else to hold the pen.
I read everything. Come argue with me.