Anti-money laundering has been an issue for banks and financial institutions for some time. Transaction monitoring systems have been around for many years. Meeting and complying with regulations at the regional or national level, however, remains a challenge. Many financial institutions are almost playing catch-up with regulators, as anti-money laundering regulations change and develop. This just-in-time compliance poses its own risks, as well as being stressful for the institutions concerned. 

At the same time, fraudsters and criminals are not standing still. Money laundering techniques are becoming increasingly sophisticated. It is perhaps inevitable that fraudsters and criminals are routinely ahead of the banks and regulators. This does not mean, however, that financial institutions can afford to stop monitoring, or even take a slightly more relaxed approach.

A vicious cycle of cost

Unfortunately, however, staying on top of changes increases the number of alerts in the system. Each one has to be investigated, but that takes time and effort. This, in turn, increases operational costs. What’s worse, standard anti-money laundering detection techniques, such as transaction monitoring and filtering, tend to result in a lot of false positives. Some estimates have put this at 90 percent or more. The number of reports to regulators suggests that this means that banks could be investigating thousands of false positives at any one time.

This is a huge drain on resources. Banks, however, are caught in a vicious cycle because failure to detect money laundering would have a major knock-on effect both financially – in terms of fines from regulators – and on reputation.

The real problem, however, is that this process is not static. Regulatory requirements increase, and so does reputational risk. Fraudsters continue to add new techniques and methods and hone the old ones. The number of alerts increases, possibly exponentially. Banks are forced into a position where the operational costs may simply be too great, and they are forced to pull out of that particular market. This, naturally, has effects on overall revenue, and they may not be going in the desired direction.

Finding a better way

The answer does not lie in behaving like a hamster on a wheel running ever faster in the hope of catching up with criminals. There has to be a better way, and there is. Banks can understand and manage risk more effectively through the use of analytics, and particularly machine learning techniques. The real value of analytics is that it can be used to reduce alerts and false positives in the system. It can, therefore, both increase the effectiveness of anti-money laundering activity and reduce operational costs.

The starting point is to examine and filter transactions differently. The first step is to segment transactions: group them together by characteristics, using analytical models. You can then apply tailored filtering to each stratified segment or group. This already reduces the false positive rate and immediately starts to reduce operational costs significantly but without – and this is crucial – increasing your reputational risk. Indeed, the actual level of risk is likely to go down. Investigating fewer cases, but in a more targeted way, means that you are more likely to detect money laundering.

Artificial intelligence to the rescue

This step is, however, only the start. The next stage is to identify potentially problematic transactions using models based on machine learning. Trained on historical transactions, a machine learning model (which can be, for example, a neural network) can provide a score for currently analyzed transactions to identify the most suspicious for money laundering. This allows you to focus attention and resources on the riskiest transactions and customers. This further significantly reduces the false positive rate and again, allows you to reduce your operational costs still further without additional risk.

Machine learning models are helpful in anti-money laundering because they allow you to analyze data from multiple sources together and to detect nonlinear relationships and interactions between variables, creating useful scores and… Click To Tweet

Last but not least, machine learning can be supported by natural language processing. Fuzzy matching and text analytics are extremely useful in reducing false matches when talking about sanctions screening or adverse media analysis.

Moreover, taking advantage of network analytics could be the key to your success in solving issues related to Correspondent Banking and Ultimate Beneficiary Owner detection.

The bottom line

Anti-money laundering requirements and regulations are certainly not going to go away. Indeed, trends suggest that they will become more important over time because criminals are certainly not giving up and going home. Banks and other financial institutions need to become more sophisticated in their detection techniques if they are to survive and thrive. 

US regulators already encourage banks to consider “innovative approaches” including new technology – such as artificial intelligence – to enhance their AML compliance programs. Other regulators will do the same soon. 

This means that now nothing should stop us from leveraging AI techniques to improve compliance management.

To find out more, I recommend this white paper on how AI can reduce money laundering. Or you may want to discover how advanced analytics, machine learning and RPA can change the speed and quality of how you detect and prevent financial crime. To do so I recommend this information.