Agentic Intelligence for Contract Review on Snowflake

Details

Snowflake's Forward Deployed Engineering team deployed a contract review agent internally to handle its high-volume enterprise deal flow — hundreds of bespoke order forms, MSAs, and amendments per quarter. Before the system, auditors manually scanned PDFs line by line and could only review a sample of contracts; the new agent processes the full population. The agent runs in three layers: ingestion and extraction from PDFs via Snowflake AI Extract, LLM-based clause classification using Cortex AI and Cortex Agent, and Streamlit-based reporting for auditors. VP Internal Audit Amrita Kapoor described the goal as flipping the model: AI handles exhaustive reading, experts handle exception judgment. The tool is entirely Snowflake-native — Openflow, Cortex AI, AI Extract, Streamlit, Snowflake Intelligence — making this a product showcase as much as a case study. The 70% reduction in review time is the headline, but the shift from sample-based to 100% population audit coverage is arguably the bigger operational win for compliance teams. Worth noting: this is a self-published case study from Snowflake's own blog. The company has clear incentives to make results look compelling, and no independent validation of the 70% claim is cited.