AI creates efficiencies in sanctions checking
The growing use of artificial intelligence (AI) and machine learning could dramatically reduce the amount of time and resources spent on sanctions checking.
In transaction banking, the focus on technological development has centred on the possibilities of blockchain technology.
However, this has overshadowed the arrival of AI into transaction-banking platforms. AI and machine learning are helping to further reduce manual checks and processes.
The first target for implementation is sanctions and compliance. As companies become increasingly international, irrespective of size, checking against sanctions has become an essential activity for more than just the MNCs. AI can learn through experience what can pass through the sanctions filter, and what compliance obligations need to be checked.
Tristan Blampied, senior product manager at Pelican, says: “Machine learning can facilitate sanctions compliance efficiency and decreases the false positive rate. It can learn from historical behaviour and experience.
“If a party is flagged multiple times, but consistently allowed through by the operator, the machine will learn to no longer flag it in future, as long as all of the contextual details remain the same. The learning process is controlled and set within the tolerance threshold of the organization in question.”
The technology can continually learn from past decisions, whether they were made by a human or the software itself, based on the context. Blampied says this means there is a constant improvement to efficiency and potentially further cost savings.
He adds that Pelican’s transaction-banking product also includes natural language processing, which can understand text even if it is written in a free format. It can then change or replace information within a payment message to allow the transaction to be completed with the minimum input. To date, the technology is able to recognize text in English, French and German.
He says: “For sanctions it is possible to scan unstructured and free-format text from financial instruments and documents and, if required convert to industry message formats such as Swift. This offers significant value over and above the simple matching of words on a list against structured information, which is not in itself an overly complex process for the technology.”
Tim Brew, CGI
Tim Brew, director, financial services at CGI, explains that employing AI can reduce manual checking requirements by 70%.
“Once implemented, banks can actually tighten their thresholds – the settings in fraud management systems that determine how close a match needs to be in order to create a positive hit – because the workload has been reduced, thereby allowing the scanning to be even stronger,” he says.
The potential for the software is not limited to these areas.
Gene Neyer, head of product management at D+H, says: “AI can really be used in almost any process. In sanctions and compliance, it is used to minimize false positives, in payment execution it can be used to improve straight through processing through electronic repair, in receivables it can be used to improve the matching of credits to invoices.”
Other industries that see a high volume of payments sent and received, such as insurance, could also benefit.
It has a further application: assessing the accumulated data flows on different algorithms to identify unusual transaction patterns, according to Brew. These could be picked up more easily than at present if they fall outside the remit of the current product solutions and silos.
CGI's Brew says: “By using big data feeds, the software can be run across the whole banking infrastructure. By developing specific algorithms and sharing knowledge, the self-learning platform can identify and prioritize good guys which when shared across the silos will drive greater efficiencies.
"The big data element means the banks can drive efficiency gains and adds additional protection.”
Implementing the technology does not necessarily mean having to overhaul the current banking software.
D+H's Neyer says: “Typically, AI is plugged into the existing software. Less typically – but potentially with a larger impact – an entirely new platform is deployed – with the existing operating platforms consuming its services.”
Although it can improve the functions, there remains a small element of risk involving transactions slipping through. However, Brew believes that the time saved through manual checking can be applied to other areas of higher priority.
He says developing new processes is a requirement in the changing landscape.
“By improving the efficiency of the whole area within a bank, it enables the institution to focus their resources on tackling and fixing fraud once identified. By bringing in new tools, such as the big-data enablers across the different silos of the bank, the chances of catching and identifying fraud is greater.
"Unlike many areas of the business, fighting financial crime is a constantly moving target, as the criminals are always trying to stay one step ahead – thus the ability to be flexible and fleet of foot is essential.”