Classification

Classification is a Databricks ai / ml capability, introduced March 2026.

Agent Bricks capability with a no-code UI that sorts documents and unstructured text into categories you define, built on the ai_classify SQL function.

• The builder previews on just five rows at a time - it "classifies up to 5 documents and rows and displays the results" - so you tune labels on a tiny sample with thumbs-up/down feedback, then scale out via SQL or a Lakeflow pipeline. 🔗
• It leans on the ai_classify SQL function, which reached general availability in June 2026 with label descriptions, global instructions, and multi-label classification. 🔗

⚠ Limitations: Built on ai_classify, which isn't available on Databricks SQL Classic and can't be used with views. Labels must number between 2 and 500 unique values, with names of 1-100 characters and descriptions of 0-1000 characters, and total context is capped at 128000 tokens. 🔗

Category

AI / ML

Introduced

March 2026🔗

Announced at

Public Preview March 2026; ai_classify GA June 2026🔗

Also known as

Text Classification, Agent Bricks Classification

Verified

2026-08-20