One thing I’ve noticed while working with enterprise analytics systems:
The hardest problems are rarely technical.
Most modern models can generate SQL.
Most warehouses are well documented.
Most organizations have catalogs and governance programs.
Yet teams still depend heavily on a handful of experienced engineers.
Why?
Because analytics depends on knowledge, not just data.
For example:
A schema may tell you that three customer tables exist.
It doesn’t tell you:
· which one is authoritative
· which one is historical
· which one executive reporting relies on
Experienced engineers know the difference.
AI doesn’t.
That’s why many enterprise analytics failures aren’t caused by bad SQL generation.
They’re caused by missing organizational knowledge.
As AI adoption accelerates, I think we’re going to spend less time talking about prompts and more time talking about knowledge infrastructure.
Because models can’t use knowledge that organizations never captured.
Top comments (0)