The EU demands that the search for money laundering and terrorist financing must be intensified. For financial service providers, this means that they have to process anomalies in payment transactions efficiently. This is where new techniques and machine learning come into play.
The global Finance Action Task Force (FATF) only tightened the measures against money laundering and terrorist financing at the end of April 2022. This has direct consequences for financial institutions such as banks and insurance companies: their checks have to be much more stringent, which means that they have to revise the current applications and, in the worst case, replace them. This is the only way they can ensure that they discover evidence of illegal payment transactions.
One of the required measures is payment screening, which must be carried out for every single money transaction and which prevents money laundering. Until now, it was sufficient to compare data from incoming and outgoing payment transactions against sanctions lists, embargo lists and other black lists. Banks usually check last names, first names, company names, aliases, domiciles, alternative spellings and posting texts.
Check payment transactions with fuzzy matching
But that’s over in this form now. In order to be compliant, banks must now be able to check names for similarities in payment transactions. The search technology known as “Fuzzy Matching” is helpful here, which also includes phonetically similar terms in its fuzzy search and also listens closely to similar words, synonyms and substitute words typical of the scene. However, the use of this search technology does not pose any major problems for banks at first, because there are corresponding modules and implementations that use official PEP, sanctions and black lists for the comparison and work with many languages, including exotic ones.
Limit the growing volume of investigation cases
The big issue is the increasing number of abnormalities to be examined manually, which a fuzzy search inevitably generates. However, the previous approach of simply assigning additional employees for compliance work is uneconomical and, given the lack of specialists, simply not feasible. It is therefore urgently necessary to develop a fundamentally different approach.
The right technical solution offers software with artificial intelligence. It relies on machine learning algorithms. This works very reliably, above all, because the software can be trained with a large number of cases in which the result of the manual check by the compliance department is already known. The AI software therefore checks the significantly larger number of hits generated by the fuzzy search and only forwards those cases to the compliance staff that are most likely to be money laundering.
In practice, this means that even fewer processes end up in the compliance department, but that they can uncover more illegal processes overall than before. By feeding the results of the manual check by the compliance department back into the AI software, the number of cases to be checked manually can be reduced even further – a pleasant side effect for the compliance staff as well.
Payment transactions: Use case of a bank from Lichtenstein
The increased compliance requirements for payment transactions have also prompted VP-Bank from Vaduz (Liechtenstein) to look for a solution that meets the new anti-money laundering regulations. The bank quickly realized that it needed a method to keep the number of cases passed on to employees for review to a minimum, despite a less precise search. Effectiveness (actually finding all conspicuous processes with fuzzy matching) should be combined with the greatest possible efficiency. Only what is most likely to be a hit is checked manually.
For the implementation, it was advantageous that the bank was able to provide extensive data for training the AI model. This allowed the bank to have the effectiveness of the AI model checked by an external benchmark institute before the new software was put into operation. After the positive validation, the software module Name MatchingTransaction the Actico Compliance Suite be used in practice. In the introductory phase, the bank also included frequent false positive cases in white lists in order to minimize the total number of hits. The proportion of valid hits in the total number of incidents to be checked manually increased accordingly.
Reduce costs with fuzzy matching and AI
The use of fuzzy matching search algorithms and artificial intelligence is a win for everyone involved: the banks can reduce the manual effort and thus their costs. The compliance employees have less to do with false reports and can focus on the difficult cases. The government and its taxpayers are losing fewer illicit money transfers that would otherwise go undetected. The fact that the system can be retrained with new findings for continuous improvement is another incentive for AI software like this Actico Compliance Suite to put.
About the author: Thomas Knöpfler is Co-Founder and Head of Compliance Solutions at Actico. He is responsible for the strategy of the Business Unit Compliance at Actico. After studying business administration, he began his career in an international consulting company. Even then, his goal was to optimize the collaboration between business departments and IT with technical solutions. He has been Managing Director since 2015 and Head of Compliance Solutions at Actico GmbH since March 2022.
Actico is a provider of intelligent automation and digital decision making solutions. The scalable software combines control technology with machine learning and is audit-proof throughout. This enables companies to process large amounts of data and make and automate AI-supported and rule-based decisions in real time. (sg)
Also read: Supply chain financing: Esker expands portfolio with equity investment in LSQ
Lead image: Edler von Rabenstein – Adobe Stock
#Payment #transactions #machine #learning #enables #efficient #payment #screening