At the Data + AI Summit, Databricks launched Agent Bricks, a breakthrough platform that enables businesses to develop and deploy AI agents using their own data—without complex tooling or manual tuning. Now in beta, Agent Bricks transforms high-level task descriptions into production-ready AI agents with full domain awareness.
CEO Ali Ghodsi called it “a whole new way of building and deploying AI agents that can reason on your data,” empowering enterprises to move from idea to execution with unprecedented speed, quality control, and cost efficiency.
Built on Mosaic AI research, Agent Bricks generates domain-specific synthetic data and task benchmarks to evaluate performance. It then runs multiple optimizations, allowing users to select the ideal version balancing accuracy and cost.
The platform supports diverse enterprise use cases—from information extraction (PDFs to structured data) to knowledge assistants and multi-agent supervisors that manage workflows like compliance checks. A Custom LLM Agent enables brand-aligned content generation at scale.
Early adopters like AstraZeneca processed over 400,000 clinical trial documents without writing code. Flo Health enhanced medical AI accuracy while ensuring privacy and safety compliance.
Databricks also launched serverless GPU support and MLflow 3.0, enhancing performance, model monitoring, and lifecycle management for generative AI.
Together, these innovations position Databricks as a comprehensive platform for building and scaling enterprise AI systems.