At RTL Nederland, we recently migrated our entire data ingestion platform — pipelines, dbt models, orchestration, and tests — spanning multiple repositories and teams. Rather than doing it all by hand, we brought AI into the workflow as a genuine collaborator: one tool to help us write and transform code, and an autonomous agent to review every pull request across the full stack.

This talk walks through what that looked like in practice, where AI genuinely saved us time, where it fell short, and what we learned about building an effective human–AI development workflow for large-scale data engineering projects.