Speaker, "Cutting Through The AI Hype Cycle: A Survey of the Current State, Industry Headwinds, and Next Steps in Transaction Monitoring," ABA/ABA Financial Crime Confernce, Oct'24

The presentation examines the current state of artificial intelligence in transaction monitoring (TM) and AML compliance, arguing that industry adoption has lagged despite widespread discussion of AI’s potential.

It distinguishes between traditional AI/ML models used for predictive analytics and generative AI tools that assist with drafting investigation narratives and SARs. Practical use cases include ML-driven transaction monitoring systems that reduce false positives and improve alert quality, ML overlays that risk-rate or auto-close low-value alerts, and generative AI “copilots” that streamline investigative workflows. While these tools offer measurable efficiency gains and broader risk coverage, challenges remain around explainability, model risk management, computational bias, and regulatory scrutiny.

The primary barriers to adoption are not technical but structural and organizational. Legacy systems, data silos, poor data quality, regulatory hesitancy, and limited board-level prioritization impede deployment. Effective AI models require feature-rich, consolidated data environments and outcome-based testing rather than rigid rule validation. Regulators have increasingly encouraged responsible innovation, signaling openness to AI within existing model risk management frameworks. The presentation concludes with practical steps for implementation, including consolidating AML data, establishing technology sandboxes, inventorying use cases, and pursuing both top-down governance and bottom-up pilot deployments to drive incremental adoption.