Data platform and AI agents
The data architecture for Kraft Heinz, with AI agents on top so business teams can ask their data questions in plain language. An early write-up that grows as the work ships.
Built with
- Snowflake
- dbt
- Cortex Agents
- Cortex Analyst
- Cortex Search
- Semantic views
The goal
Business questions should be answered from one governed model of the data, not from spreadsheets that disagree. The warehouse is designed first and the agents are built on top of it, so every answer traces back to the same definitions.
How the data will flow
Business systemsthe sources the business already runs on
dbt star schemasfacts and dimensions
Semantic viewsone definition per metric
AI agents on Snowflake CortexAnalyst answers from tables, Search answers from documents
Business teamsthe people who ask the questions
Beyond the build
- Use cases: AI use cases are evaluated for business value, and the viable ones are shipped.
- People: requirements from business and technical teams are turned into working solutions.
Client data, schemas and internal details are not shown.