Databricks expands Agent Bricks and releases new AI functions in public preview

Databricks has expanded its Agent Bricks page to cover all Databricks AI capabilities. The page previously focused on declarative agents and now acts as the central point for the full range of AI features. Two new user interface experiences for document analysis are now in public preview. The Classification experience provides a visual interface for assigning documents to predefined categories. Users can configure labels, preview results on samples, and then apply the process at scale. The Information Extraction experience offers a visual interface for pulling structured data from documents, tables, PDFs, and images. It supports defining and refining extraction schemas with preview options before operations at scale. Both interfaces are built on v2 signatures of the ai_classify and ai_extract functions. These AI functions now support v2 signatures in public preview with improved quality and performance along with greater flexibility. The signatures also enable native composability with ai_parse_document to power end-to-end document processing workflows.