Scaling AML compliance with AI to prepare for the next black swan event

Black swan events, by definition, don’t happen often, but when they do their effects can present significant and unique challenges to compliance programs at banks and other financial institutions (FIs).

For example, after Russia’s invasion of Ukraine in February 2022, anti-money laundering (AML) and sanctions compliance organizations, for the most part, were not well prepared for the breadth and depth of the additional work needed to maintain effectiveness in their programs. It’s safe to say we haven’t seen a sanctions environment change so drastically or as rapidly in a very long time, if ever. The breadth of the sanctions, as well as the coordinated effort from all the global regulatory agencies were unprecedented. Moreover, to compound the effects, it wasn’t all done in one fell swoop. Rather, there were sanctions released in batches — sometimes every few days, other times perhaps every week or two. And as soon as organizations would get a handle on one batch of sanctions, it would start all over again. This presented a vicious cycle, stressing organizations’ compliance teams.

Keeping up with sanctions and politically exposed persons (PEP) lists are challenging enough under normal conditions. Being able to scale your operations for the next black swan event – or even the increasing volume surge as real time payments become ubiquitous – is critical to an effective and efficient program.

This is where artificial intelligence (AI) can help innovate, scale and transform organizations’ AML compliance programs.

Drowning in false positives

The manual review of vast volumes of data, a cornerstone of traditional compliance practices, has become increasingly labor-intensive, time-consuming, and error prone – with 99% of the alerts being false positives. This alert fatigue can overwhelm analysts and hinder their ability to focus on genuine threats. The risk of missing a true positive or escalation becomes more real as screening teams become stressed with volumes where nearly all alerts are false positives. In other words, human analysts become conditioned to expect all the alerts to be false positive, thereby missing the true positive.

The result is that many of today’s AML and Sanctions compliance programs need not only a boost but also a heave.

Organizations have typically sought to meet sanctions compliance needs by throwing bodies at the problem. However, there aren’t enough skilled people, particularly for level one (L1) teams, to handle the heavy workloads, not to mention the repetitive processes of monitoring transactions and reviewing alerts that can lead to employee burnout.

We’re seeing organizations in a perpetual hiring challenge. Hiring and onboarding new analysts and often retraining them can be a lengthy process and fraught with challenges, including an increased risk of errors and missed escalations. Unfortunately, turning to outsourcing, offshoring, or temporary employment is not making much of a dent in the problem. Plus, organizations end up competing for the same resources, whether they be contractors or managed service providers.

Typical sanctions alert review analysts can work 200–300 alerts per day at high levels of quality. When alert volumes spike an individual analyst would need to work upwards of 500–800 alerts per day. Handling that many alerts per day is not sustainable for any analyst. Analysts who strive to keep pace with alerts rush their work and eventually start to make mistakes. Others simply give up and robotically click through alerts and provide dispositions that are based on inadequate reviews, enabling criminal activities to go unnoticed. Yet, to staff up and meet the new volume levels would require a FI to increase analyst head count by approximately 100%-200%. That’s prohibitively expensive and time consuming.

What’s required is a more comprehensive approach that allows for more effective risk management, improved quality, and scalability — an approach that helps solve the “people issue” that has persisted for the past two decades. AI, combined with people, can help organizations better allocate resources to manage increasing and fluctuating workloads, as well as allow analysts to focus on higher risk alerts and higher value work.

Need for transformation

Using time and money to review thousands of false positives is an efficiency problem that can lead to an effectiveness problem. Analysts may miss the bad actor “needle in the haystack,” that rare true positive, due to resource strain from reviewing thousands of false positives, where they expect every alert to be a false positive. Implementing AI tools can help banks and FIs automate repetitive tasks, enhance decision-making capabilities, and scale their capacity, thereby increasing the effectiveness of the program.

Not only is AI fast, but it is also accurate. Fewer manual touchpoints mean less risk for human error. And AI can differentiate between genuine risks and false alarms to ensure that genuine threats receive prompt attention.

Implementing AI into your compliance programs: crawl, walk, run

The journey toward AI-enabled AML compliance begins with a strategic approach to implementation. Banks and FIs must assess their existing processes and identify areas ripe for automation and optimization.

I always recommend starting small. Identify a single use case to deploy quickly. This will help you get familiar with how you would approach AI and ML within your program. Not all AI is created equal, and there are challenges to keep in mind:

1. Ensure the transparency and interpretability of AI models. This will help to understand how the AI made decisions and ensure compliance with regulatory requirements.
2. Prioritize data privacy and security. This is ever important to safeguard sensitive information and mitigate the risk of unauthorized access.
3. Validate AI models thoroughly. Ensure accuracy, reliability, and consistency in detecting suspicious activities and minimizing false positives.
4. Address potential biases in AI algorithms. Prevent discriminatory outcomes and ensure fairness in decision-making processes.
5. Monitoring and oversight. Ongoing efforts are essential to evaluate the effectiveness of AI-driven compliance solutions.

Banks and FIs will always need people. By designing a volume-independent compliance program that can easily scale, you can automatically deliver more effective, robust, timely, and cost-efficient compliance. And, by automating routine tasks such as sanctions screening alert review, employees can focus their expertise on more high-value activities such as risk assessment and investigation.

Tomorrow’s AML and sanctions compliance will need a combination of people and AI working smarter together.

Art Mueller is Vice president of Financial Crime at WorkFusion.