Association of Certified Anti-Money Laundering Specialists (ACAMS), the largest international membership organisation for anti-financial crime professionals, has released a Best Practice Guide to Transaction Monitoring(TM). Entitled Effectiveness Matters, it argues that only true collaboration between public and private organisations will unleash the crime-fighting prowess of Artificial Intelligence (AI)enabled TM systems.
“Eventually, governments and law enforcement agencies need to work together with policy makers to agree on top risks and direct the private sector to focus efforts towards detecting those top risks,” claims the guide, which includes contributions from a range of AML experts including Napier’s Chief Product Officer, Will Monk.
The guide examines the impact of regulatory technology, from the introduction of perpetual Know Your Customer (pKYC) to the move towards AI and Machine Learning (ML), on the ability of Financial Institutions (FIs) to identify the risks they couldn’t pinpoint using legacy systems. It outlines how organisations have successfully blended the old rules-based systems with newly available technology.
Attendees at the round tables staged by ACAMS to discuss ways forward in best practice cited their major TM challenges as poor data quality, poor integration across systems and information islands, inhibiting organisations from being able to take a holistic approach, or to accurately measure the impact of new technology. Limited interaction with regulators and law enforcement agencies also adversely affects an understanding of how to produce actionable Suspicious Activity Reports (SARs).
As a result of these issues, some FIs are sticking to the safety of rules-based systems, while others are introducing ML to augment their systems. While it is broadly agreed by all that AI is the future for TM, cost, implementation timescales and explainability still represent barriers to its full-scale adoption.
Despite the challenges, participants in ACAMS’s discussion felt that AI, with its unparalleled ability to accelerate the detection of thematic risks, is the future – but how soon the future arrives is still under discussion.
Building a robust Transaction Monitoring system
The report recommends designing a solution to meet a defined user need; building in a strong governance and model validation process including explainability and audit; and including a feedback loop exploring client and organisation-wide risk.
For now, rule-based systems will continue to function in tandem with AI until FIs and regulators become comfortable with AI. However, such systems can be enhanced, network analysis tools incorporated, systems tuned, data quality improved and processes streamlined to optimise results.
The incorporation of ML, using well trained algorithms and good quality data, enables FIs to deploy human resources to investigate higher risk alerts, redirecting their efforts from spending time clearing a huge number of false positives. ML can also be used to enhance the model and fine-tune its setting for achieving optimal results.
While AI is not the only way forward yet, given the trajectory of adoption by regulators and larger organisations, it is only matter of time before its adoption is commonplace. However, it can only be introduced when its outcomes are fully explainable and backed by strong risk-based justification. AI solutions are able to demonstrate the effectiveness of individual rules, identify risks that have been going undetected, identify riskier behaviour patterns, and help achieve that elusive balance between effectiveness and efficiency, transforming Transaction Monitoring outcomes.
Getting it right – how to achieve successful transformation
The guide recommends seeing vendor selection through the lens of a multi-year journey, encompassing data cleansing and consolidation, algorithm training, pilot programmes and implementation. While investment in data reconciliation maybe be considerable, FIs which have undertaken the process reported better controls across the anti-financial crime ecosystem. The most successful transitions engaged regulators at an early stage, to understand and incorporate their expectations and comments. In some instances, the adoption of an AI based solution resulted in an increase in alert volumes because previously undetected risks came to light.
Few FIs interviewed by ACAMS could give a clear outline of how they measured effectiveness, but continually using product risk assessment and finetuning to ensure that controls always cover the whole ecosystem, irrespective of mergers, acquisitions and legacy systems. Other effectiveness measures cited were applying governance to the release of new products at the outset,and ensuring that KYC data was robust.
Alert conversion rates shared varied from 3% (97% false positives) to 20% (80% false positives), but industry sector was a key determinant in these variations.
The guide suggests that alert conversion rates are at best blunt instruments in measuring efficiency, with a more accurate measure being the effective direction of human interventions.
“The guide is a great resource for any FI considering upgrading their Transaction Monitoring solution,” says Napier’s Will Monk. “Its advice is practical, pragmatic and based on the lived experience of experts and compliance teams.”
The full version of the Best Practice Guide is available to ACAMS’s Enterprise customers, and the Executive Summary to non-Enterprise customers.
Access your copy here
Read the Napier teams’ 5 key takeaways from ACAMS Hollywood 2023
Over the past year, AML and anti-financial crime experts have encountered a multitude of unexpected crises and emerging risks on a near daily basis. New challenges call for new solutions. To that end, the Napier team joined thought leaders from the public and private sectors as well as NGOs, for the annual ACAMS Hollywood conference. Read our top 5 key takeaways here
Photo by Henrik Dønnestad on Unsplash
.png)
