For two years, we built on this insight — one nobody cared about. We did. That belief is two years ahead of the market that just caught up.
Then came February 3rd.
Snowflake announced Semantic View Autopilot at BUILD London, a tool that scans query history to automatically build semantic layers. Thirty-plus organizations have joined the Open Semantic Interchange initiative. Gartner has called it essential infrastructure.
The market has collectively decided that AI without business context doesn't work.
But here's what's still missing
The semantic layer is supposed to be the meaning layer between your raw data and the people — or AI — trying to use it. The thing that knows "active patient" means last visit within 90 days. That "revenue" excludes trials and one-time charges. That "Northeast region" is a specific list of states, not a geographic guess.
You can't build a meaning layer without the meaning. And the meaning isn't in your data catalog. It's not in your Confluence pages. It's not in anyone's head in a way that scales.
It's encoded in the SQL your analysts have been writing for years — thousands of queries, years of decisions, hiding in plain sight.
- Not in your data catalog
- Metadata tools capture structure, not meaning or intent.
- Not in Confluence
- Documentation is incomplete, outdated, and unsearchable at scale.
- Not in anyone's head
- Tribal knowledge doesn't scale and walks out the door with every departure.
- It's in your SQL
- Thousands of queries, years of decisions — hiding in plain sight.
Consensus Without Visibility Isn't Governance. It's Guessing
Snowflake's Autopilot picks the consensus definition. That sounds right until you realize: before you can trust any definition, you need to see the disagreement.
- When does a deal count as closed?
- At signature, at payment, at go-live?
- Is a customer churned after 30 days of inactivity or 90?
- Finance, product, and commercial teams may each have a different answer.
- Does "treatment naive" mean no prior therapy ever, or just none in the last 12 months?
- In regulated industries, a wrong cohort definition can make it into a submission that shouldn't.
In most organizations, finance, product, and commercial teams are each calculating these differently. Nobody knows it until something breaks — a board number that doesn't match a dashboard, an AI output nobody trusts, or in regulated industries, a cohort definition that makes it into a submission that shouldn't.
And there's a problem nobody else is even talking about
Everyone — Snowflake included — is focused on governing what's in your warehouse: the tables, the columns, the schema. But a huge amount of your most critical business logic never makes it to a table. It gets calculated in a query, exported to a CSV, emailed to leadership, and it's gone. The metric existed for one moment and then vanished. Nobody tracked it. Nobody versioned it. When the analyst who wrote it leaves, the logic leaves with them.
Snowflake is building semantic layers on top of your schema. We're surfacing the logic that never made it into the schema in the first place. — and why we believe query labeling is the unlock.
We mine your query history — and surface what no tool has shown you before
Across Snowflake, Databricks, BigQuery, Redshift, or any SQL environment:
Your real metric definitions
Every meaningful business rule encoded in your queries, structured and searchable — not buried in logic someone wrote three years ago and nobody has touched since.
Where your definitions conflict
The places where the same term means different things to different teams, so you can resolve the disagreement before it becomes a bad decision or a compliance problem.
Your key person risk
The analyst whose departure would quietly break things for months before anyone figured out why — because 80% of your most critical logic lives in their queries and nowhere else.
The knowledge is already there. The only question is whether you find it first
In 4 to 6 weeks, you get a dashboard of all of it. Your team makes the decisions about what to preserve. We make the invisible visible.
- Organizations that move now
- Will have a mapped, trusted knowledge foundation before their competitors have finished arguing about whose revenue number is right.
- The ones that wait
- Will spend the next year watching their best analysts leave — and their institutional knowledge leave with them.
The knowledge is already there. The only question is whether you find it first.