JPMorgan Chase's COIN platform saves 360,000 legal and loan-officer hours annually with AI contract analysis
The bank's Contract Intelligence system reviews commercial-loan agreements in seconds, eliminating the manual workload that once consumed 360,000 hours per year across legal and lending operations.
5 min read
Hours Saved/Year360000 hrs
Background
Commercial-loan agreements require intensive manual review for interpretation and compliance. Prior to COIN, this work consumed 360,000 hours annually from legal and lending staff at JPMorgan Chase.
What Was Implemented
- Machine learning and NLP platform (COIN) trained to read and interpret commercial-loan agreements
- Processing of approximately 12,000 new wholesale contracts per year
- Automated identification of key provisions, reducing reliance on manual review by lawyers and loan officers
- Go-live: June 2016
Results
COIN reduced the annual review workload from 360,000 hours to seconds per document. The system also cut loan-servicing mistakes attributable to human error. Bloomberg and the ABA Journal independently corroborated these figures from JPMorgan's own disclosures.
Lessons
- AI contract review delivers most value when it removes low-cognition, high-volume document processing from skilled professionals
- The framing of "freeing people for higher-value work" (vs. "replacing people") drove adoption
- Combining ML with NLP enables interpretation of legal language, not just keyword matching