Is context engineering the new analytics engineering?
Checkr built a structured context layer on top of its dbt project, and it let an AI agent trace a support-metric miss in minutes instead of hours.
Overview
Analytics engineers built transformations to define and share metrics, but AI agents need more than definitions—they need context. Covers Checkr's context engineering approach using dbt and Omni: guardrails for AI, LLM-assisted tooling, and feedback loops that improve AI performance over time.
What you'll take away
- 01
Connecting an AI agent directly to a 6,000-model Snowflake and dbt project produced a different wrong answer every time the same question was asked — a pattern Checkr calls metric drift.
- 02
AI-readiness breaks into three layers: the warehouse (the dictionary), the semantic layer (the grammar that prevents metric drift), and AI context, the conversational bridge most teams skip entirely.
- 03
With synonyms, SLA targets, and edge-case rules encoded, an agent cut an FRT investigation from hours to minutes and correctly ruled out the queue everyone assumed was the culprit.
Session speakers
Chris Merrick
OmniSarah Fischbach
Checkr