The fight against money laundering is a huge challenge for financial institutions around the world, and as criminal methodologies grow more sophisticated, so must the AML measures put in place to stop them. Since modern AML requires firms to deal with vast amounts of complex customer data, many are turning to technology, specifically artificial intelligence (AI) and machine learning (ML) systems, to help them detect money laundering activities and satisfy their evolving compliance obligations.
AI holds such promise as an AML tool because it not only performs AML tasks faster than a human compliance employee but, via machine learning, has the capability to adapt to new threats and new money laundering methodologies, ensuring that firms are able to re-position quickly in different regulatory environments and stay one step ahead of the criminals.
What is AML Artificial Intelligence and Machine Learning?
AI, deployed as part of a firm’s AML program, represents a series of algorithms that control the digital measures put in place to detect money laundering (and other criminal activities). These algorithms analyze huge amounts of customer data, including customer due diligence (CDD), sanctions screening and transaction monitoring inputs, to perform a variety of automated tasks in order to remediate suspicious activity.
The application of machine learning within an AI infrastructure means that AML programs can, in theory, become even more efficient. By utilizing previously analyzed CDD and transaction monitoring data, machine learning tools can gauge emerging customer behavior and make a more accurate determination as to the level of money laundering risk that behavior represents.
The practical advantages of ML and AI money laundering tools to an AML program are as follows:
Changes in Behavior:
When customers’ transaction data is inputted into an AML program, machine learning models can analyze that behavior to make predictions and judgments about that customer in the future. More specifically, with the benefit of ML, an AML system could become sensitive to changes in behavior, however subtle, that conventional AML checks might miss. Those deviations from the norm would represent a new set of data inputs that could, in turn, be analyzed by an AI algorithm to determine whether a suspicious activity report (SAR) is warranted.
Automated AI systems allow the CDD and Know Your Customer (KYC) processes to take place faster and with greater depth and scope. The quality and quantity of AI-enhanced CDD will give compliance employees a greater range of relevant AML data that can be used to inform risk assessments, suspicious activity reports and subsequent investigations. In more detail, the CDD and KYC applications of AI will allow firms to:
- Efficiently collect identifying data from a greater range of external sources, including sanctions lists and watch lists, in order to construct a more accurate customer risk profile.
- Identify beneficial owners of customer entities in the same manner, using external data faster and more efficiently.
- Aggregate and reconcile customer data across internal systems to eliminate duplication and errors and enhance the consistency of AML measures between customers.
- Automatically enrich suspicious activity reports with relevant data from customer risk profiles or data from external sources.
Beyond building customer risk profiles, AML compliance requires the analysis of unstructured data as part of transaction monitoring, PEP screening, sanctions screening and adverse media monitoring processes. In order to properly assess the risk those customers present, firms must attempt to use that data to understand their social, professional and political lives, examining a range of external sources, including media and public archives, social networks and other relevant datasets.
AI systems help firms manage and analyze that unstructured data in a way that enhances AML compliance. In practice, that means running AI-assisted customer name searches against vast amounts of external data and finding matches, patterns and connections that would be missed by other types of conventional analysis. Once the data is collected and analyzed, AI can help firms prioritize and categorize information to aid risk management.
Suspicious Activity Reporting:
Artificial intelligence can aid suspicious activity reporting (SAR) by not only generating reports automatically but automatically filling them out with relevant information. Prior to their submission to the authorities, SARs often go through an internal reporting process with contributions from numerous AML employees and senior management. The internal process may even require the submission of data from different parts of the world and in different languages.
AI technology can be used to make the SAR process easier: algorithms can pre-populate automated reports with relevant data and present that data with accessible, standardized language and terminology in order to minimize bureaucratic friction and ensure consistency for every contributor. By standardizing language and terminology, and so increasing the narrative focus on regulation, AI can not only increase the speed and efficiency of a firm’s AML reporting, but also its impact in subsequent investigations by authorities.
The principal advantage of technological automation within an AML system is to add speed and efficiency to otherwise complex and time-consuming compliance procedures. However, one of the major obstacles to compliance efficiency, even in an era when technology allows firms to perform AML with greater speed, is the level of noise, or false positives, that result from incomplete or inadequate data and the over-sensitivity of AML measures.
In practice, thanks to false positives, only a fraction of AML alerts progress to become full SARs: a rate that implies a high degree of wasted time, money and resources. With that in mind, AI and ML systems promise a significant transformative effect to the level of noise generated during the AML process: AI can help firms generate much richer insight into customers and transaction patterns, allowing them to eliminate incorrect and irrelevant alerts that make the compliance process so costly for firms and onerous for customers. Practically, those applications include:
- Semantic analysis of alerts to identify those created by redundant data.
- Statistical analysis of high-risk customers and transactions to distinguish true positives from false positives.
- Intuitive screening during sanctions, PEP and adverse media checks to eliminate mistakes and false positives generated by regional naming conventions.
- Prioritization of higher risk customers during the transaction monitoring process.
It’s worth bearing in mind that the strengthening relationship between AI and financial crime compliance will not eliminate the need for human AML teams or the development of risk-based AML programs specific to their environments. By reducing noise, AI and machine learning tools allow AML employees to better prioritize and address the most urgent money laundering alerts with the benefit of human experience and expertise and, in doing so, more effectively contribute to the fight against financial crime.