← All updates Databricks Update One-Pager Navy version
databricks Platform · Updated Jul 15, 2026

AI Extract and Classify Functions

Enhanced AI capabilities for compliance-focused workspaces.

What it does

The `ai_extract` and `ai_classify` SQL functions are now enabled by default in workspaces with the compliance security profile activated. `ai_extract` allows users to extract structured fields from unstructured text based on a user-defined schema, while `ai_classify` categorizes text into predefined labels for efficient document management. These functions support various compliance standards, enhancing their applicability in regulated environments.

Why it matters

This update significantly improves the ability of organizations to manage and classify unstructured data in compliance-heavy industries, thereby enhancing productivity and reducing the risk of non-compliance. By automating the extraction and classification processes, businesses can streamline workflows, improve data governance, and ensure adherence to regulatory requirements.

Key capabilities
Structured Data Extraction
Use `ai_extract` to pull structured fields from unstructured text.
Text Classification
Utilize `ai_classify` to categorize text into user-defined labels.
Agent Creation
Create information extraction and classification agents for automated workflows.
Compliance Support
Functions support multiple compliance standards including HIPAA and PCI-DSS.
Prerequisites
Serverless compute enabled
Unity Catalog enabled
Access to a serverless usage policy with a nonzero budget
Unstructured data in a Unity Catalog volume or table
Recommended use cases
Classifying customer call transcripts by intentCategorizing documents by content typeTagging product reviews by sentiment
Limitations · remember
!Function availability may vary by region
!Requires at least one unlabeled document or row in Unity Catalog to build an agent
!Must have at least two labels defined to run classification
Architecture
Compliance Security Profile → Unity Catalog → AI Functions → Classification Agent
Get started
1
Create Agent
Go to Agents and select Create Agent for Text Classification.
2
Select Source Data
Choose a volume with documents or a table with text data.
3
Configure Labels
Add and refine classification labels for your data.
4
Run Classification
Save and run the classification to see results.
5
Use Agent
Deploy the agent for large-scale classification tasks.
Key takeaway
The availability of AI extract and classify functions by default enhances data management capabilities for compliance-focused organizations.
2026-07-15-cc15310b Generated by the Databricks Updates Agent · browse all updates