Databricks Agent Bricks powers enterprise agents

The interesting part is not that Databricks has an agent builder; it’s that it’s trying to make agents feel like a data platform primitive instead of a one-off app. That’s the right shape for enterprises that care more about control, evaluation, and routing than flashy demos.

• –Agent Bricks centers the workflow on enterprise data, which matters for document parsing, citations, and repeatable outputs
• –The platform framing reduces glue code by bundling optimization, evaluation, and deployment into one system
• –Pairing with GPT-5.5 signals a multi-model enterprise stack rather than a single-vendor model story
• –This is most compelling for teams already living in Databricks, Unity Catalog, and governed data pipelines
• –The risk is abstraction creep: if the workflow gets too opinionated, power users may still drop to custom frameworks for edge cases