Meta:Context: a business context schema in dbt's semantic layer
Green Dot built a five-layer context schema inside dbt's semantic layer so AI agents know not just the number, but what it means.
Overview
Addresses how AI can better understand enterprise data through business context: a 5-layer, 36-field schema encoding business and metrics knowledge in dbt's existing meta: block. Shows smaller language models paired with structured metadata can match larger models' analytical reasoning. Covers schema design, theoretical foundations, simulation evidence, and starting with a single metric.
What you'll take away
- 01
A semantic layer gets the query and the math right but doesn't tell an agent what a number means; Green Dot's Meta:Context schema adds five typed categories, descriptions, expectations, investigation, relationships, and decisions, right in the metric's own YAML.
- 02
Because context lives in the same file as the metric, it goes through the same review process, pull requests and CI, and one API call returns both the number and the reasoning behind it.
- 03
Asking seven different AI models to write a context card from raw materials found all seven flagged an untested claim appropriately, but none reconstructed the one rule that came from an analyst's actual investigation.
Session speaker
Keith Binkly
Green Dot Corporation