Dollar bills

Banks, machine learning, and the war against money laundering

Money laundering is the term for deliberately masking the origin of illegally obtained cash generated by criminal activity. Money laundering activity is on the rise, fueling the operations of drug cartels and terrorist organizations around the globe. Advances in technology allow so-called “megabyte money” to exist on computer screens and move anywhere in the world nearly instantaneously. According to data on money laundering from the United Nations, 2% to 5% of global GDP is laundered annually, or between $800 billion and $2 trillion dollars.

Banks and other financial services companies are on the front line of anti-money laundering (AML) efforts. AML regulations require financial institutions to monitor deposits and wire transfers to spot signs of suspicious activity. An American Bankers Association survey of AML enforcement at banks with assets larger than $1 billion showed that 30% of banks intend to increase AML team headcount and budgets, citing the need for new software to remain efficient and the overall increase in suspicious activities that require review.

Research from McKinsey highlights the operational challenges banks face in ensuring effective and efficient AML compliance. They write:

AML compliance programs now look more like operational utilities or, as one executive put it, “factories,” and less like the independent oversight functions that banks first envisioned. These factories are expensive, yet might be acceptable if the huge teams and manual processes were working well. But many are not.

One central challenge to AML compliance is that banks’ legacy monitoring systems are often rules-based and inefficient, with data showing that as many as 90% of AML alerts are false positives. Meaning that a lot of effort and expense goes into investigating legal, compliant wire transfers and deposits. AI-powered solutions are a clear need, but advanced automation carries its own challenges.

A range of factors contribute to the difficulty of automating AML efforts. Low-quality data and fragmented data sources make automated solutions difficult to develop. New banking products and services, like instant fund transfers and mobile payments, create more platforms that need to be monitored. And the inconsistent availability and quality of data across geographies makes standardizing processes difficult.

Despite the challenges, banks see machine learning and data analytics technologies as the future of AML fraud detection. These systems have the potential to reduce the manual work of compliance personnel by 50%, effectively doubling their effectiveness dollar for dollar. The potential financial impact of these AI systems could be huge. Using the UN estimate of global money laundering, a 50% cut in illegal activity would represent a $400 billion to $1 trillion decline in dirty money transfers. Which would create a positive real-world impact slowing the flow of money to illegal organizations worldwide.