JPMorgan Chase expanded an internal large-language-model pipeline to middle-market commercial lending on Tuesday, telling investors the system processes financial statements, tax filings, and covenant histories for companies seeking credit up to $50 million. The bank reported a 34 percent reduction in time from application to term sheet for standardized industries.
Chief operating officer Daniel Pinto said the models do not set interest rates or approve loans autonomously. Relationship managers receive draft memos highlighting leverage trends, customer concentration, and peer benchmarks, then must sign attestations before committees review deals.
Regulatory Scrutiny
The Office of the Comptroller of the Currency requested documentation on bias testing after community advocates flagged disparate error rates in extracting data from smaller firms without audited GAAP statements. JPMorgan said it retrained parsers on 12,000 anonymized loan files and engaged an external auditor.
Competitive Pressure
Goldman Sachs and Bank of America run parallel pilots. Regional banks without comparable AI budgets risk slower turnaround times that corporate treasurers increasingly cite in RFP decisions.
Employee Impact
The bank's union for operations analysts sought retraining guarantees. Management committed $40 million to internal courses on model oversight and credit policy, though headcount targets for the lending division remain unchanged this year.
What Borrowers See
Clients notice faster preliminary feedback, not different pricing algorithms exposed to them. Lawyers advising mid-cap issuers recommend reviewing consent language for AI-assisted document processing in credit agreements.
