Machine learning is currently on the tip of everyone’s tongue across all industries. Particularly, there is a wide variety of use cases in the financial industry. Thomas Ohlemacher, product manager at ACTICO GmbH, explains in an interview which options this technology offers for compliance management in banks.
BANKING NEWS: The ACTICO Compliance Suite has been supporting banks and insurance companies for more than 15 years. How have the challenges in compliance departments at banks changed over this time?
Thomas Ohlemacher: During this period, banks and insurance companies have been subject to one wave of regulation after the other. Each wave brought about the urgent need to fulfill new requirements. At ACTICO, this led to the development of a Compliance Suite with wide-ranging coverage of compliance issues, spanning AML to counter-terrorism financing, PEP check and KYC, through to market abuse, insider trading and fraud. Banks and insurance companies are continuing to switch to this solution, but the motivation for doing so has changed in recent years. It is less a matter of covering new requirements, but rather that of adopting established standards and becoming more efficient. In doing so, more and more new technologies, such as machine learning, are being used.
Machine learning and artificial intelligence are often mentioned in the same breath. How do these terms differ from each other?
Artificial intelligence encompasses a wide range of approaches. Machine learning is one of these approaches and is currently of particular relevance in practice. It has been used successfully in fields outside compliance for some time. In the neighboring area of fraud prevention, machine learning is used to learn from old fraud cases, while it is used in marketing to identify sales opportunities based on data.
Which use cases does machine learning offer for compliance management at banks?
At ACTICO, machine learning complements the existing standard models for monitoring and rating. Allow me to describe two specific use cases. When monitoring payments, machine learning can learn from existing payment histories. It recognizes patterns that have conspicuous payments in common. Since machine learning can create very powerful models, it is often possible to filter payments more accurately and trigger fewer unnecessary queries. As a second example, I would like to mention the comparison of names against sanction and PEP lists. In this case, machine learning learns, from the examination history, how the involved parties queried historical cases. This model then makes it possible to evaluate future cases. In this way, you get an early estimation of whether a hit is likely to be relevant and you can prioritize accordingly.
What preparations does a bank need to make before implementing a machine learning software platform?
In the field of compliance, surprisingly few preparations are necessary. These have often already been largely fulfilled by the introduction of a compliance solution such as the Compliance Suite. With this solution, the data histories that are required for machine learning have already been collected.
What is the implementation cost and how long does the training on the machine learning models last?
It varies from case to case. But we already managed to train some interested parties on the first helpful models within only a few days. This enabled them to quickly evaluate the results themselves.
Do compliance managers need advanced technical know how to work with this platform?
The compliance managers do not need any technical know how. Some banks and insurance companies already have their own data scientists who deal with the models. Others have the models created by ACTICO. The effectiveness is demonstrated to the compliance managers through statistics and transparent test cases.
Source: Banking Club (German)
Find out more in the detailed white paper “Why Successful Banks Apply Machine Learning to Compliance Now”, which you can download for free here.