Oct 28, 2026: AI_
Snowflake Cortex AI_PARSE_DOCUMENT introduces AGENTIC mode in Public Preview, turning complex enterprise documents—from financial reports to technical manuals—into actionable information for AI applications, analytics, and automation. Go beyond text extraction to unlock information in charts, diagrams, tables, and other visuals with a single SQL call, without building or managing complex document-processing pipelines.
Key use cases include:
- Intelligent document processing: Transform documents containing text, tables, and visuals into usable information for business processes and AI applications.
- Agentic document workflows: Automate document interpretation and processing, reducing manual effort and custom engineering.
- Multimodal retrieval-augmented generation (RAG): Make information embedded in charts, diagrams, and screenshots accessible to enterprise search and RAG applications.
- Structured data extraction: Extract chart series and values, table headers and cells, and diagram components and relationships for analytics and automation.
- Enterprise knowledge discovery: Uncover insights from reports, presentations, and manuals by making both textual and visual information accessible to search and AI applications.
AGENTIC mode identifies and classifies visual content, selects and executes processing actions, and generates semantic summaries, structured data, or both. Snowflake manages the orchestration, returning document text, layout, visual summaries, and extracted data in a unified response.
For more information, see Agentic mode.