Context layer for AI agents
Give AI agents the context to answer data questions correctly: definitions, metrics, join paths and access rules.
Outcome: You can design the context an agent needs to turn a business question into a correct, governed answer.
- 1
What context an agent needs
A schema tells an agent where data is stored. To answer correctly it also needs definitions, metrics, join paths, rules and freshness.
Build concepts and relationships · 12 min · instant checks
- 2
Descriptions agents can act on
An agent picks a metric by reading its description. Say the grain, what is included and what is excluded.
Define governed metrics · 12 min · instant checks
- 3
Synonyms and disambiguation
People say Customer, Client and Account for one thing, and Account for two different things. Write both down.
Build concepts and relationships · 12 min · instant checks
- 4
Governed join paths
Relationships with specific verbs are the paths an agent walks to write joins. Declare them, and the agent stops inventing its own.
Build concepts and relationships · 12 min · instant checks
- 5
Guardrails: PII and what agents must not see
Mark personal data, move it to a restricted table, and keep the tables agents query free of it.
Build it in the Model Builder · 14 min · instant checks