Machine Learning (ML) and Artificial Intelligence (AI) are often touted as the way forward to enhance Anti Money Laundering (AML) and to successfully apply Know Your Customer (KYC) standards. As many banks seek to further strengthen their AML systems, it’s important to understand the key role that technological innovation plays.
This blog is the first in a series of short articles and events on technological innovation in AML, as part of a knowledge partnership between the European Banking Federation (EBF) and SAS. SAS is a global thought leader and solutions provider in this field. EBF has been keen on contributing to an effective regulatory framework and to interdisciplinary cooperation in relation to AML in Europe, including through calling on legislators to create an enabling environment for the employment of new technologies in the AML/ Counter Financing Terrorism (CFT) domain. The knowledge partnership between EBF and SAS will provide the banking sector and its institutional stakeholders with new insights and, through multi-stakeholder dialogue, will help to explore and promote the implementation of new solutions.
The current state of affairs
In December 2020 SAS, in partnership with ACAMS and KPMG, surveyed more than 850 compliance professionals and ACAMS members from across the globe. The results were published in the “ACCELERATION THROUGH ADVERSITY, The state of AI and ML adoption in Anti Money Laundering compliance” paper.
The findings present a snapshot of AI and ML adoption, its challenges, and potential untapped opportunities. One key takeaway from the survey was that AI and ML are gaining serious momentum in AML compliance. More specifically, the survey proved that the three primary benefits of AI and ML adoption in AML compliance are:
- Improved quality of investigations and regulatory filings.
- Fewer false positives and lower operational costs.
- Better detection of complex risks.
The key challenge
One of the most critical tasks banks face in complying with Anti Money Laundering (AML) regulations involves detecting suspicious activities which include looking for possible terrorist financing activities (CTF). In a previous blog post, SAS discussed key techniques that can be used to improve AML strategies. Banks in practice rely on AML systems to scan transactions in search of anything suspicious. Often, these are rules-based systems that look for common money laundering patterns which then generate alerts requiring the intervention of AML investigators. Investigators determine if the alert is a false positive or something that needs to be escalated and reported as suspicious to the relevant financial intelligence unit (FIU).
But money laundering activities are complex. And money launderers routinely attempt to evade detection by creating sophisticated transactional patterns that blend into the crowd of genuine transactions. It’s difficult for traditional rules or scenario-based AML systems to spot these sophisticated data patterns. Compounding the issue, business scenarios that perform well today won’t necessarily perform well tomorrow. As a result, banks see a high volume of false-positive alerts generated by outdated scenarios or valid scenarios running against ever-changing data. This is where the use of more advanced technological solutions comes in: AI and ML can, for instance, help reduce false-positive alerts in a way that stays abreast of the development of new business scenarios.
Preparing for AI and ML adoption
Before embarking on a journey towards Machine learning and AI it is critical to set the stage for success, by evaluating your current AML transaction monitoring system. This initial stage involves understanding how well each of your business scenarios performs. This requires defining and applying a consistent measure across all business scenarios – which varies from bank to bank. For example, you can define measures as alerts that:
- Are filed as suspicious activity reports (SARs) or suspicious transaction reports (STRs), and confirmed as suspicious by the regulator.
- Are filed as SARs or STRs.
- Have certain disposition codes (e.g., closed as true-positive).
- Required more than X number of hours or days to investigate.
It’s essential for all parties to understand and align to this definition as the information gathered from this process will help you prepare for the use of AI and Machine Learning. The knowledge gained in this phase is key for a successful Machine Learning and AI innovation project.
Before going into the use of machine learning and AI, it is also highly recommended to improve your current AML transaction monitoring effectiveness. Once you’ve established the process for defining and measuring the performance of your business scenarios, you’re ready to apply three techniques to improve the effectiveness of your AML transaction monitoring system.
SAS explored in detail how the three techniques of segmentation, threshold setting or tuning, and alert hibernation can help to optimize your current AML Transaction Monitoring system.
The conditions and preparations described above for the use of ML and AI in AML activities of banks will be discussed in more detail during a series of masterclasses that EBF and SAS will conduct in the first quarter of 2022. The first two masterclasses will focus on
- Innovation with Machine Learning and AI to enhance AML Transaction Monitoring.
- AML Data Governance and Lineage as it relates to KYC
You can find more information on these masterclasses and register here. Meanwhile, stay tuned for our next blogs during 2022 and be part of our conversation.
Follow us on Twitter @EBFeu