Within the financial services sector, Anti-Money Laundering (AML) is a significant challenge for many institutions, often consuming large numbers of people and effort to manage the process and comply with the regulations. As a result, these same institutions are looking for new solutions to help them reduce the burden and increase the controls in this complex space. The combination of artificial intelligence (AI) and, more specifically, machine learning (ML), are increasingly being considered as enablers of a better solution.
Despite its potential, however, adoption of AI and ML within Anti-Money Laundering has been relatively slow. This is due, in part, to the limited understanding of how AI and ML could be applied within compliance programs, and to the fact that regulators and compliance officers are often concerned that AI and ML are "black boxes" whose inner workings are not clearly understood. Regulators typically require compliance officers to understand and validate not just the outputs, but also how the outcomes from AML models are derived. Despite some of the concerns, we already see movement and application of these technologies.
Machine learning has been shown to be particularly useful in conducting suspicious activity monitoring and transaction monitoring, two key AML activities. A common challenge in transaction monitoring, for example, is the generation of a vast number of alerts, which in turn requires operation teams to triage and process the alerts. ML can teach computers to detect and recognize suspicious behavior and to classify alerts as being of high, medium or lower risk. Applying rules to these alert classifications can facilitate the automatic closing of alerts, allowing humans to supervise the machines that triage these alerts rather than reviewing all of the alerts manually, and making better use of the time of these experts.
Institutions leveraging ML can reduce their dependency on human operators to perform routine tasks, reduce the total time it takes to triage alerts, and allow personnel to focus on more valuable and complex activities. There will always be a need for human involvement in the AML process; in fact, hybrid human/AI models and processes are the direction we see the function moving towards and should enable AML transaction monitoring to take a step forward in both the efficiency and effectiveness of alert operations teams.
To implement ML as part of a transaction monitoring solution, firms need to get key elements in place. These include:
â?¢ High quality data. All monitoring systems and analytics, not just ML applications, depend upon high quality data. Static files such as Know Your Customer data as well as dynamic data on customer transactions held by financial services firms frequently have low completeness ratios in areas such as payment information, along with high error rates. Profile refreshes, conducted as part of sales and marketing exercises, can update data while increasing customer outreach and identifying cross-selling opportunities.
â?¢ A 360-degree view of the customer. Currently, financial services firms do not have the global freedom to share information about their customers to build a comprehensive network, and they do not formally collaborate on AML initiatives. Regulators are, however, increasingly leaning toward data sharing between banks. Over time, as ownership and privacy concerns are addressed, large amounts of transactional data could become available on intrabank data clouds, making a 360-degree view of the customer more feasible.
â?¢ Expertise in financial services and ML. Very few people are experts in both ML techniques and financial services. As a result, there have been fewer applications targeting financial services problems from start-ups and established vendors, limiting acceptance of ML within the sector. Firms hiring ML experts can provide the needed financial expertise, if they institute appropriate training and development programs.
â?¢ Straightforward systems and processes. ML is a relatively new technology and there are few established, straightforward processes to follow to implement it. Without knowing what to look for, teaching systems to detect certain types of financial crime can be tricky. For example, how does one teach a system to recognize terrorist financing? There are more established processes for managing fraud, but nothing as comprehensive for terrorist financing, other than name matching against terrorist lists.
Financial services firms are making progress in addressing these challenges and their appetite for automation is increasing rapidly. Many banks have started implementing business process automation in the form of Robotic Process Automation (RPA). In fact, robotics and AI/ML solutions can exist independently of each other and each can support the other's capabilities. Robotics can be used to train AI/ML models and AI/ML models can be used to add decision-making or reading capabilities to robotics models.
In Anti-Money Laundering, as in so many other areas of compliance, operations, risk and finance, AI and ML could be important steps in financial services firms' journey to greater efficiency and effectiveness. These improvements in compliance and resilience capabilities can help to benefit firms' shareholders and customers, make regulators' jobs easier, and strengthen the global financial system