AML Transaction Monitoring: rules-based system efficiency, machine learning and agentic AI

> Client Risk Rating, Client Segmentation, Rules (Selection, Parametrisation & Tuning)
> Community detection, graph-based representations, detection of anomalies

Machine Learning: useful techniques for Transaction Monitoring

Machine learning for improved transaction monitoring systems.

• Community detection leads to the segmentation of clients.
• Graph-based representations uncover hidden relationships.
• Feature engineering creates detailed representations of customers.
• AI for the detection of anomalies identifies money laundering rings.

The three components of an efficient AML TM rules-based system and the usage of Machine Learning

Client Risk Rating

• Goal of CRR, typical risk factors to be considered and examples of rule-based approaches.
• Data sources, triggers and interaction with Compliance processes.
• Testing and assessment: key methodologies.

Client Segmentation

• The benefits of Client Segmentation and differences between a segmentation for business and financial crime purposes.
• Top-down segmentation driven by business input.
• Data-driven bottom-up client segmentation.
• Analytics to test the client segmentation.

Rules: Selection, Parametrisation & Tuning

• Rule based monitoring: from simple to complex rules.
• Parameter selection and tuning for efficient AML TM systems and why the input from client risk rating and segmentation is important.
• Above and below the line testing: validation based on statistical samples.
• Machine Learning models as an enhancement or complementary solution to rule-based systems.

Future Trends

Agentic AI for AML Compliance

• What is Agentic AI?
• How can it benefit the AML function and what are its limitations?
• Discussion of existing use cases and key factors for success.
• Outlook to future developments.