From dbt Core to Fusion: How CarGurus unlocked dbt State and AI at scale

Migrating from dbt Core to the dbt platform was the central element of a complete reimagining of CarGurus' Data Engineering stack. This session covers how we led our 80-person developer team through the full migration: moving from GitHub Server to GitHub Enterprise Cloud, replacing Airflow with dbt platform orchestration, and upgrading our codebase to the dbt Fusion engine—the foundation that made everything else possible. Fusion unlocked cost optimization and AI workflows at scale. dbt State alone cut Snowflake compute costs by 25% and delivered ROI in under 5 months by reprocessing only impacted models. Self-service enablement reclaimed 24,000 working hours in a year. We'll also cover governed AI tooling: the dbt MCP server, agent skills for PR reviews and data triage, and workflows spanning Snowflake, Atlassian, Omni, Glean, and Slack. Attendees will leave with a step-by-step migration playbook and a framework for layering AI on a modern dbt foundation.