How AI for Financial Decision Making Works: A Guide for the US Financial Market

How AI for Financial Decision Making Works: A Guide for the US Financial Market describes how banks in the United States are deploying AI while complying with longstanding model-risk rules. A risk officer at a US regional bank asked a newly hired AI lead how to use a large language model to summarize loan committee memos without violating the same model risk framework that applies to credit scorecards. The answer, reportedly 26 pages, referenced Federal Reserve guidance from 2011 and the joint SR 11-7 model risk management approach with the OCC. The article explains that major banks typically run multiple model layers, including classical machine learning, deep learning, large language models, and reinforcement learning. It notes that top institutions maintain model inventories of roughly 2,000 to 4,000 active models, reviewed by federal examiners during annual supervisory cycles. For production AI, documentation and independent validation—including methods like red-teaming and bias audits—are required, with validation teams adapting to large language models’ complexity.