AI won't skip the boring data governance work

Many people seemed to hope that with AI we could finally skip the "boring" data governance work. All those policies, definitions, and documentation. Just plug in some LLMs, add a few AI agents, and magically get instant insights and natural language queries, all by itself.

But reality kicks in: the things we tried to ignore for many years are back as non-negotiable prerequisites. Without clean, documented, well-structured data, AI won't do any magic. In fact, the fastest way to get AI delivering value isn't adding more sophisticated AI - it's good old data governance.

It reminds me of many Data Science projects about a decade ago that quietly turned into data engineering projects along the way. Same as now, ML and models were not the critical part, but the data was.