Ship AI agents like tables: dbt materializations for Snowflake CoWork

At Under Armour, we deploy Snowflake Intelligence agents Cortex Agents, Semantic Views, and Cortex Search Services entirely through dbt Cloud. We wrote custom dbt materializations that let us ship AI objects the same way we ship tables: through pull requests, code review, and environment promotion. 

Every agent has a PR history. RBAC grants run through post-hooks. dbt's DAG traces lineage from raw data through transformed models to the agents sitting on top. When a semantic view changes, we know which agents are affected before we merge. This talk covers how we built the custom materializations, the create-or-alter deployment logic, and how our data team owns the full AI layer without a separate deployment process. You'll leave with a practical pattern for making AI deployments reproducible, auditable, and CI/CD-native using dbt.