COiN reads commercial loan agreements in seconds, not 360,000 hours a year
JPMorgan Chase · 2017–ongoing
What they did
JPMorgan's Contract Intelligence platform (COiN) uses NLP to extract clauses, dates, and obligations from commercial credit agreements — the document-review grind that previously ate law-firm and in-house-counsel hours by the tens of thousands.
What happened
The bank reports COiN reviews roughly 12,000 commercial credit agreements a year in seconds each, recovering an estimated 360,000 hours of lawyer and loan-officer time annually — about $150,000 in annual review time reclaimed for every agreement type it's pointed at, at scale. It has since expanded toward additional document types beyond its original loan-agreement scope.
The so-what
The oldest, least glamorous AI win in enterprise finance is still the most reliable one: narrow, high-volume, structured-output extraction from documents a human would otherwise read line by line. No agent loop required.