Richard Shearer, CEO of Tintra PLC
Let’s imagine a scenario in which an individual living in Kenya wants to send money to London. If the person were living in the UK, this would be a very simple matter: two taps on their smartphone phone would result in a near-instantaneous transaction with no questions asked.
As a cross-border payment, however, this transaction looks very different: The individual’s money will have multiple arduous hurdles to clear as it’s passed from a local bank to a local electronic money institution [EMI] to a UK EMI, then on to a UK bank, before – if they are lucky – arriving in the hands of the beneficiary. This is a long-winded process riddled with red tape that could last as long as three months.
In addition, this already lengthy process is compounded by a further impediment: because the person is from an emerging country, they will – in the eyes of KYC/AML compliance teams – be considered high risk, despite the fact their earnings and transactions are entirely above board.
In essence, these challenges boil down to two key and interrelated issues: compliance processes are full of friction [often in the form of sluggish, manual, complex, and time-consuming KYC checks from multiple banks and EMIs], and Western KYC/AML teams are subject to bias, meaning they can’t [or possibly won’t] discern ‘good’ clients from ‘bad.’
Therefore, in order to truly achieve global banking, we need to develop a solution that will allow financial institutions to deal with both challenges quickly, efficiently and automatically – which is where technology, and more specifically, artificial intelligence, has the answer.
Leveraging technology to eliminate barriers and address bias
The banking industry hardly needs persuading of the benefits of AI in broad terms, as demonstrated by a recent report from McKinsey,which signposted several advantages of its integration, including boosted revenues, lower costs, and the discovery of unrealised opportunities through insights generated by powerful, data-hungry technology.
Perhaps more importantly, however, AI has the potential to significantly improve AML processes. For example, in predictive analytics, machine learning methods can be used alongside customer data to predict possible criminal behaviour – at lightning speeds and at an unprecedented scale – which simply cannot be matched by the people who remain at the heart of legacy banks’ compliance teams.
Not only can AI speed up the cumbersome processes that create the kinds of barriers faced by the likes of the individual from Kenya, but – crucially – the adoption of AI can also help to overcome the significant barriers represented by KYC/AML bias.
For example, using cutting edge AI tools to streamline onboarding and compliance procedures and automate all processes that currently involve manual invention, will effectively replace subjective human decision making with intelligent machines that have learned from years of data and experience. As a result, by reducing human involvement to a minimum, these tasks become fast, fair, transparent, scalable, and flexible enough to be applicable to customers and transactions across the globe.
Of course, AI isn’t always entirely free from bias – it’s made by people, and its insights are interpreted by people too. This reinforced by the last Nordics Anti-Financial Crime Symposium, which highlighted the need to watch out for bias at the programming stage.
In the context of KYC/AML classifiers, an unfair bias could occur if the machine is trained to mimic the human decision-making process, where the ‘right decision’ is fed into the AI solution. This can be overcome by providing evidentiary data instead, where the machine can learn from examples of transactions that resulted in complications as opposed to modelling outcomes on potential human prejudice.
Another key challenge for AI is generalisation caused by ‘narrow’ training data, such as when certain demographics and/or ethnic groups aren’t represented sufficiently in the training set. A similar phenomenon can occur in the context of KYC / AML where criteria for accepting a customer or transaction can vary across geographic area, meaning those in emerging markets may suffer as a result.
That said, it doesn’t mean AI can’t help in eliminating prejudice in AML procedures – far from it – it simply means we need to ensure the next generation of fintechs and challenger banks utilising this technology are feeding their AI models good data that provide explainable results – and that these entities are sincere in their desires to level the global banking playing field.
Revolutionising the global finance industry
Taking this kind of technology seriously would be nothing short of revolutionary for the global finance industry.
After all, as the Centre for Global Development has recently noted, KYC/AML discrimination can have serious ramifications in emerging markets, with those most likely to be impacted including “the families of migrant workers, small businesses that need to access working capital or trade finance, and recipients of life-saving aid in active-conflict, post-conflict, or post-disaster situations.”
In looking beyond the benefits that this new breed of global banking will have on individuals, there are also huge implications for the global economy.
McKinsey’s report on the future of cross-border payments points out that international payments revenues already amount to around $200bn globally – but a closer look at the figures reveals that while Western Europe sees 5.5 annual cross-border transactions per capita, Latin America only sees 0.7.
If compliance barriers were lowered through the leveraging of new technology, it seems perfectly plausible to suggest that places like Latin America would see cross-border transactions increase, with all the economic benefits associated with this increased flow of money on an international scale.
And, with AI and machine learning leading the charge towards revolutionised banking, it’s worth remembering that decreased prejudice needn’t come at the cost of increased risk: in fact, a recent Deloitte survey found that 41 per cent of respondents believed too many false positive AML alerts were the biggest AML compliance challenge faced by banks today.
Therefore, the right technology operated by new, forward-thinking financial entities has the real potential to simultaneously address the prejudices that underpin AML compliance processes, eliminate the sluggishness that those processes entail, unlock new streams of money to circulate in the global economy, and address the current lacklustre state of addressing financial crime.
When one really allows oneself to really absorb this new paradigm, the potential is there for AI to completely repackage the way in which the global banking industry operates. The question is who will be first to the party!
