Extraction Agent
Pull structured records from dense documents
Extraction Agent
Due diligence requires reading thousands of pages just to find key dates and numbers. The Extraction Agent pulls this data into structured grids instantly.
Setting Up Extractions
You define what you are looking for by creating Extractions — individual questions, each with a name, a data type (Boolean, Date, Number, or Text), and optional instructions for the AI. For example, you might create extractions for "Target Company", "Liability Cap", and "Governing Law".
When you feed 150 contracts into the Extraction Agent during a Routine, it will populate a clean matrix of answers across all 150 documents. Each answer includes a Confidence Score and a direct citation to the exact paragraph where the data was found.
If the Confidence Score drops below 80% (often due to ambiguous phrasing in the contract), the system will flag that specific cell for manual human review, ensuring you don't miss nuanced risks.