What AI Taught Me About Enterprise Data

I used to think the hardest part of analytics was collecting data.
Then I thought it was data quality.
Lately, I've started to believe it's something else entirely.
Relationships.
The longer I work with enterprise systems, the more I realize that most of the important knowledge isn't stored inside tables.
It's stored between tables.
Experienced engineers know this intuitively.
They know which joins are safe.
They know which mappings should never be used.
They know which paths produce the numbers executives trust.
The problem is that this knowledge rarely gets documented.
AI has a way of exposing hidden problems.
And one thing it's exposing right now is how dependent many organizations are on relationship knowledge that exists only in people's heads.
Maybe the next step in data infrastructure isn't collecting more data.
Maybe it's preserving more relationship knowledge.