Governed by default: how data teams at Nordstrom turn dbt governance into an AI advantage
Nordstrom data leaders show how a governed data foundation on dbt keeps an AI assistant from confidently handing back the wrong number.
Overview
Explores how Nordstrom turned dbt from a pipeline tool into a control mechanism for reliable AI: treating data freshness as an AI SLA, using tests and contracts as safeguards, and leveraging lineage for transparency. Demonstrates "Data Concierge," a production conversational agent answering business questions in natural language using governed data, with real production decisions and tradeoffs.
What you'll take away
- 01
Governance was the AI unlock itself, not a compliance afterthought: Nordstrom built a multi-agent "data development lifecycle" that generates pull requests, contracts, and tests as hard gates in a medallion pipeline.
- 02
Onboarding to dbt only five months before the talk, Nordstrom already saw pipeline builds that used to take weeks ship in days, with freshness running as an automated, testable requirement.
- 03
Their advice for teams starting out: put tests, contracts, and lineage in place early since they become the control surface for whatever AI gets layered on top, and centralize semantic definitions once rather than re-solving them per interface.
Session speakers
Priya Tanwar
NordstromNadine Bruxel
Nordstrom