Information Extraction

Information Extraction is a Databricks ai / ml capability, introduced June 2025.

Agent Bricks capability with a no-code UI that turns unstructured documents into a structured table by extracting fields against a schema.

• You describe what you want in plain English and click Generate Schema - it writes the extraction schema for you; under the hood it runs on the ai_extract SQL function. 🔗
• The original launch capability now sits at a '(legacy)' docs page, replaced by a rebuild on the ai_extract v2 signature. 🔗

⚠ Limitations: Information Extraction agents have a 128k-token maximum context length, and union schema types aren't supported. 🔗

Category

AI / ML

Introduced

June 2025🔗

Announced at

Data + AI Summit, June 2025 (beta); rebuilt on the ai_extract v2 signature, Public Preview March 2026.🔗

Also known as

Key Information Extraction

Verified

2026-07-23