Ship AI Agents Like Tables: dbt Materializations for Snowflake CoWork
Under Armour turned Snowflake AI agents into ordinary dbt models, so every agent now goes through the same pull request, review, and CI process as every table.
Overview
Under Armour deploys Snowflake Intelligence agents through dbt Cloud using custom materializations, managing AI objects (Cortex Agents, Semantic Views, Cortex Search Services) through the same PR/review/promotion workflow as tables, with full audit trails, RBAC via post-hooks, and DAG lineage from raw data through AI agents.
What you'll take away
- 01
An AI agent should ship the same way a table does: Under Armour wrote custom dbt materializations so agents, search services, and semantic views live as ordinary dbt models with full lineage.
- 02
A single command shows every agent a change would touch before merging, turning a bad change from a manual scramble into a simple git revert.
- 03
Five months after go-live: 30-plus Cortex objects deployed through dbt across three domains, with same-day answers replacing overnight reports for the commercial team.
Session speakers
Stanley Bathina
Under ArmourRyan Ballman
Under Armour