Data governance isn't failing because it's too slow. It's failing because we keep building the roof before the foundation.

In the data warehouse era, we drowned in siloed reports. In the big data era, we drowned in lakes no one could navigate. Now, in the AI era, we're scaling chaos at a speed that would have been unimaginable a decade ago.

And most organizations still aren't ready for it. I know — I lived it firsthand inside the Government of Canada.

The ones quietly winning with AI right now aren't the ones with the most data. They're the ones who did the unglamorous work first — clear ownership, consistent definitions, a single source of truth. The boring stuff. The stuff that doesn't make for a good conference keynote.

Instead, what I watched unfold was a masterclass in avoiding the hard work. Endless metadata strategies. Tagging frameworks. Taxonomy working groups. Governance committees reviewing the governance committee's terms of reference. Meanwhile the actual foundations? Still crumbling.

Tagging your chaos doesn't govern it. It just makes it easier to find.

And then AI showed up and rewarded every shortcut taken over the past decade — all at once.

Roll out Microsoft Copilot without getting your RBACs and group policies under control first, and you haven't given your team a productivity tool — you've handed it the keys to everything. Sensitive HR files. Financial records. Executive communications. Suddenly accessible to anyone who knows how to type a prompt. But hey, at least the metadata was tagged correctly.

Getting the foundations right means four things:

Accountability before automation. Know who owns each data asset before you let a model act on it.

Access control before access sprawl. Lock down your RBACs and group policies early. AI will inherit whatever permissions exist — good or bad.

Definitions before dashboards. If two teams can't agree on what "active customer" means, no governance layer will stop your AI from confidently producing two different answers.

Trust before scale. Prove data quality in one domain first. Then expand. The compounding returns are worth it.

The organizations that treat data foundations as a strategic investment — not an IT cost center — will be the ones still standing when the dust settles on the current AI wave.

The shortcuts always cost more in the end. Trust me — I have the battle scars to prove it.

But hey, at least the metadata was tagged correctly.

Originally published by Ross Norrie, founder of SkyeConnex, on LinkedIn.

Published March 19, 2026 · More from the SkyeConnex blog