The useful question is what changes for users, developers or buyers, and whether the announcement stays industry context or becomes something people can actually use.

The U.S. Food and Drug Administration (FDA) oversees products accounting for 20 cents of every consumer dollar, generating vast regulatory data. To manage this, the agency’s Office of Digital Transformation developed ELSA, a generative AI platform, and Halo, a governed data foundation running on Databricks, accessible to all 16,000 FDA employees. Previously, eight independent centers—handling drugs, biologics, devices, and more—operated separate AI tools and data stores, leading to duplication and inefficiency.

The consolidation effort began after recognizing the lack of unified data access. Within three to four months, the team integrated 50 to 60 data sources from all centers into a single Databricks platform. Real-time data streaming replaced slow batch processing, and Unity Catalog provided the governance layer to secure sensitive regulatory data while enabling controlled access across the platform.

ELSA was deployed to all staff, allowing users to select from multiple AI models via a single interface. Adoption surged from less than 1% to 85% within two months. Notably, usage extended beyond basic queries, with staff across roles creating custom agents weekly. The platform’s architecture combines structured, governed data with accessible tools, enabling non-data scientists to build agents grounded in FDA-specific documents.

The platform’s impact is evident in time savings: drug reviewers previously spent days manually extracting starting materials data from millions of pages. Now, using Databricks ML and NLP via MLflow, they retrieve answers in about three minutes. The FDA plans to expand this model, adapting tools across centers to streamline regulatory reviews and focus staff on core evaluations.