AI in Finance

Finance is one of the most demanding environments for deploying AI, and a revealing stress test for modern machine learning. This talk explores how ML and GenAI systems are actually used in financial risk decisions, where models assist accountable humans, operate on noisy and delayed data, and must remain explainable and auditable long after deployment. Through real-world examples in credit risk and NLP, we show why standard ML metrics often fail to reflect decision impact and trust. These production constraints motivate a shift toward deeper, more faithful forms of interpretability, with lessons that extend beyond finance to other high‐stakes AI systems.