A curated list of public reports, regulator guidance, and case studies worth your time. Updated when something new lands.
Annual surveys on how banks adopt AI — adoption rates by function, value pools, and where productivity gains actually land. McKinsey's Financial Services Insights hub is where their banking AI work lives.
→BCG's running thesis on why most AI programs in banks stall at the pilot stage and what the ~10% that scale actually do differently. Strong on operating-model and talent gaps.
→Quarterly survey on enterprise GenAI adoption. The financial-services slices are useful for benchmarking your pace against peers in compliance, ops, and customer servicing.
→Cross-industry framing of where AI changes the structure of banking — disintermediation risks, talent migration, new entrants. Useful for executive-level conversations.
→The clearest public picture of how UK banks actually use ML today. Where models live, what governance looks like, and which use cases are most common. Survey runs every couple of years.
→The FSB's view on systemic risks from AI in finance — model concentration, third-party dependencies, opacity. Read this before you bet your stack on a single foundation-model vendor.
→Technical research on AI's effect on credit allocation, monetary-policy transmission, and bank supervision. BIS papers tend to land 12–18 months ahead of formal regulator positions.
→Staff papers and Governor speeches signal where the Fed is heading on model risk for AI systems. Especially relevant for SR 11-7 interpretation as it applies to generative and agentic models.
→Practical guidance on what an examiner will look for when you put an agent into a regulated workflow. Pair with the Fed's SR letters for the full risk-management picture.
→Cross-jurisdictional view — how Europe, North America, and Asia are diverging on AI rules in finance. Useful if you operate across borders and need to anticipate compliance fragmentation.
→Concrete deployments of Claude in financial-services and adjacent workflows. Useful for honest before/after numbers rather than the projected savings most consulting reports lean on.
→Independent academic benchmark for how fast AI capability is actually moving — model performance, cost curves, deployment patterns. The financial-services chapters cut through vendor noise.
→We're happy to compare notes on what these reports get right and where the reality on the ground differs. 30 minutes, no pitch.