Compliance Engineering & AI Assurance
Compliance is a large and rising cost for financial firms, and AI agents are arriving in it fast. At City Agentic Finance Lab (CAFiL), we engineer and test the evaluation and assurance methods that let firms and regulators rely on them.
What we explore
AI agents in compliance operations — regulatory change, Consumer Duty monitoring, complaints handling, and financial-crime compliance operations, studied for governance, accuracy and auditability; model risk management under the PRA’s supervisory statement SS1/23; evaluation and assurance methods; regulator-ready evidence; RegTech and SupTech. We do not build detection or surveillance systems.
Questions we are asking
- Where do AI agents in compliance operations work, and where do they fail?
- How should firms evaluate and assure an agent before and after it goes live?
- What evidence will supervisors expect, and how can firms produce it efficiently?
- How do the Consumer Duty and the Senior Managers & Certification Regime apply when an agent acts?
- How should model risk management extend to agentic systems?
What we produce
- Evidence study (in preparation, with MSc projects): “Compliance agents: what works, what fails”
- “The Autonomy Ladder”, a governance toolkit
- Evaluation methods and test scenarios for agents in regulated tasks