# Agentic AI for banks in the USA US banks face detailed expectations on model risk, fair lending, privacy and cyber security. We help banks and credit unions deploy AI agents that take on real operational work while staying inside those expectations, with explainable decisions and human approval where it matters. ## Key facts - **Regulators:** Model risk, Consumer protection, Privacy and cyber, AML: see below - **Deployment options:** AWS, Azure and Google Cloud US regions, government clouds, or on-premises - **First agents we build:** BSA/AML alert research, Card and transfer disputes, Credit file preparation - **Controls on every agent:** Own identity and least-privilege access, human approval for material actions, full audit trail, prompt-injection testing ## Use cases - **BSA/AML alert research:** Gather KYC, transaction history and public information for each alert and draft the narrative an investigator needs for a SAR decision. - **Card and transfer disputes:** Collect evidence, track Regulation E and Regulation Z deadlines and draft the response for an analyst to approve. - **Credit file preparation:** Assemble the file and draft a recommendation with specific reasons, so underwriters can decide and adverse action notices stay accurate. - **Commercial loan servicing:** Read borrower financials, test covenants and flag breaches for the relationship manager. - **Member and customer service:** Handle routine requests for banks and credit unions through approved systems, with a clean handover to staff. - **Treasury and cash operations:** Reconcile accounts, forecast cash positions and propose sweeps for treasury approval. ## Regulators - **Model risk:** Federal Reserve SR 11-7 and the matching OCC guidance on model risk management, which reach AI models that support decisions. - **Consumer protection:** The CFPB expects specific, accurate reasons for adverse credit decisions, including when AI is involved. - **Privacy and cyber:** GLBA safeguards, FFIEC guidance and, for New York-licensed firms, NYDFS Part 500. - **AML:** Bank Secrecy Act and FinCEN requirements for monitoring and reporting. ## FAQ ### Are AI agents subject to SR 11-7? Agents that make or support material decisions usually are. We document purpose, data, testing and limits so model risk teams can validate them. ### Can AI support credit decisions under CFPB rules? Yes, but the lender must still give specific and accurate reasons for adverse actions. We design credit-related agents so every recommendation can be explained and a person makes the final call. ### Where should a US bank start? With work that agents prepare and people approve: fraud case drafting, KYC file preparation, servicing requests or reconciliation.