Meta:Context: a business context schema in dbt's semantic layer
Breakout sessionPowering AI-ready dataIntermediateData practitionersAll industriesTechnology
AI needs business context to deliver relevant, useful insights about enterprise data -- what numbers mean, how to investigate anomalies, what to do about them, and much more -- but so far, there is no framework.
We built one. Drawing on expert works in semantics and ontology, we created a 5-layer, 36-field schema that encodes business and metrics knowledge in dbt's existing meta: block. It flows through the Semantic Layer API, requiring no new tooling.
In simulations, Haiku with structured meta context matched Opus with scattered documentation — the schema preserves meaning well enough that a cheap, small model reconstructs high-quality analytical reasoning by using it.
This session covers the schema design, expert foundations, the simulation evidence, and where to start: one metric, three fields, one question.
We built one. Drawing on expert works in semantics and ontology, we created a 5-layer, 36-field schema that encodes business and metrics knowledge in dbt's existing meta: block. It flows through the Semantic Layer API, requiring no new tooling.
In simulations, Haiku with structured meta context matched Opus with scattered documentation — the schema preserves meaning well enough that a cheap, small model reconstructs high-quality analytical reasoning by using it.
This session covers the schema design, expert foundations, the simulation evidence, and where to start: one metric, three fields, one question.
