AutoQA: An Interpretable Automation Framework for CDD Quality Assurance in Financial Services

Quality assurance (QA) of customer due diligence (CDD) in business banking is still largely manual: analysts spend hours reviewing complex onboarding decisions against evolving regulatory and internal policies, with inconsistent outcomes and high operational costs. We present AutoQA, a deployed framework that automates key elements of the QA process by combining deterministic rules with targeted LLM-assisted reasoning, standardising decisions and surfacing issues earlier while keeping analysts in the loop. The system delivers clear outcomes and concise justifications for each step in the QA process, aligning directly with bank guidance and subject matter expert (SME) input. Its reusable, versioned checks and governed orchestration enable scaling across products, entity types and review contexts. Evaluated on real cases, AutoQA achieves over 85% agreement with human judgement and a 55% reduction in review time. Deployed in production at a leading UK financial institution, it improves throughput and “right-first-time” rates, lowers rework and reduces compliance exposure. This pragmatic, interpretable design sets the foundation for progressively complementing post-hoc QA with earlier, more automated controls across the customer onboarding lifecycle.