AI helps HSBC cut false alarms from flagged payments
An AI tool designed and implemented by KPMG has enabled HSBC to save nearly a quarter of its sanctions monitoring spend. With the international bank huge fines and long-term reputational damage if it were to be found in breach of international sanctions, the AI tool allows for alerts of sanctions breaches to be assessed and dealt with more quickly and accurately than via a manual human process.
Financial sanctions are restrictive measures imposed on individuals, organisations or governments in an effort to curtail their activities and to exert pressure and political influence on them. These restrictive measures include, but are not limited to, financial sanctions, trade sanctions, restrictions on travel or civil aviation restrictions.
As the list of entities subject to such sanctions continues to grow, banks have come under mounting pressure, as they have had to check that cross-border payments do not break rules that prevent money being channelled into the hands of corrupt business leaders, politicians or countries such as Iran or North Korea. With millions of cross-border payments made each day, finding sanctions breaches presents banks with the need to find a needle in a haystack on a regular basis – as only a tiny fraction of such transactions actually break rules. The problem is, should just one of those instances slip through the net, then the consequences for banks can be dire.
As an example of this, global bank HSBC was formerly placed under a five-year Deferred Prosecution Agreement by the US Department of Justice in December 2012, after it was found to have failed to prevent unrelated money laundering offences involving Mexican drug cartels. That instance cost HSBC $1.9 billion in redress to US authorities, along with severe reputational damages.
If HSBC was to avoid a recurrence of such a disastrous event, it would need to change its set-up when it came to flagging up potential breaches. Like many banks, it traditionally relied on basic software that flag up potentially problematic payments, which are then manually checked by two bank employees. The problem was that around 95% of these alerts turned out to be false alarms once they were manually checked.
This led HSBC to contract consultants from KPMG to develop a new machine learning tool to hasten the assessments of the alerts – and improve their accuracy. The tool was built to trawl through years’ worth of HSBC’s sanctions data, and create an algorithm that would better understand what transactions were likely to break the rules. As the algorithm was self-learning, it would keep on improving as the months went by. At present, the tool already classifies 99.9% of alerts correctly, compared with 95% for the human reviewers doing an initial first-pass screening.
Speaking to Risk.net, UK Head of Financial Crime Technology at KPMG, Robert Dean said, “If you look back at the last 10 years and all the sanctions fines, most of the banks have responded by simply turning up protection, throwing more people at it and creating huge shared service centres full of people trying to clear some of the sanctions alerts they have been generating… It was all created from scratch – there was nothing on the market we could lift and shift straight away. So it was bespoke to HSBC.”
The tool eliminated 80% of the false alarms flagged by the bank’s old sanction-screening software, dramatically reducing the workload for human reviewers. This has allowed HSBC to find an overall cost saving on its sanctions monitoring spend of 25%. KPMG is currently trialling the product with a number of other clients, meanwhile, and intends to roll out the technology to other financial crime sectors, such as name screening and transaction monitoring.
