Everyone is racing to build a semantic layer for generative BI. I get it. The demos are impressive.
But here’s what’s being missed.
For years, your analysts have quietly compensated for messy data. They knew which revenue figure finance actually trusts. They knew that a store’s numbers looked off because of a system migration, and which field you should never use. They had workarounds, caveats and judgement.
None of that knowledge is written down, and your AI won’t have access to it.
Take the human out of the loop, and every gap they were covering becomes visible. Except it doesn’t look like a gap. It looks like a confident, well-written answer that’s wrong.
Before generative BI, you need:
– A metrics catalogue
– Clear metric ownership
– A metrics architecture and agreed definitions
– Critical data elements identified
– A data quality baseline and an improvement plan
– Data access management, so the AI sees only what users should
GenAI doesn’t fix your data problems. It removes the safety net.
I’ve spent a lot of time helping organisations get this sequence right. It isn’t glamorous, but it’s the difference between insight people trust and a very expensive chatbot.
If you’re on this journey, I’m happy to share what works. Drop a comment or send me a message.