From first models to AI-powered maturity: scaling data engineering with dbt at Verisk Underwriting
Verisk modernized decades of insurance data, from mainframe to dbt, to get AI-ready.
Overview
Explores evolving dbt from a basic modeling tool into a full data engineering platform at Verisk Underwriting: AI-assisted development, using lineage for faster issue resolution, choosing between dbt jobs and Snowflake Tasks, and deploying dbt Copilot for model creation and documentation.
What you'll take away
- 01
Verisk's modernization moved through four stages:, mainframe, Snowflake with notebooks, native Snowflake objects, and finally dbt, each stage solving one scaling problem and exposing the next.
- 02
dbt unified three competing sources of truth, patched production logic, lagging Git code, and stale Confluence docs, into one description written once and referenced everywhere it's needed.
- 03
That governed foundation, not a new AI tool, is what let the team layer on AI safely: propagating documentation automatically and enforcing standards on every pull request.
Session speakers
Mateusz Geca
VeriskVijay Singh
VeriskCasey Schwegel
Verisk