Elly Yates-Roberts |
This article was originally published in the Autumn 2019 issue of The Record. Subscribe for FREE here to get the next issues delivered directly to your inbox.
We are on the verge of a technology wave, which will prove to be as disruptive as the introduction of the web browser or the smartphone. This will be the maturing of the promise of artificial intelligence (AI) into a usable technology, which will soon be as familiar as Google Maps, Siri or Netflix.
Machine learning is not a new concept; in essence it is a set of techniques that enables computers to ‘learn’ patterns and rules from historical data which enable them to make automated decisions on new data. Banks have been using these techniques for years, particularly in the fight against financial crime and in offering new targeted products to customers. However, with recent technological advancements, we are now in a position to really utilise what this approach can offer to the world of payments.
Gartner estimates that by 2020, more than three-quarters of retail customer interactions will be handled by AI agents and banking is sure to follow. Machine learning may not be true AI, but many banks still consider it invaluable to their operations. Using AI with machine learning can enable banks (retail and corporate) to communicate with their customers personally and therefore, more effectively. This approach would be far more targeted. Big data and machine learning can help banks regain control by demonstrating that they truly understand their customers. They can, therefore, adapt selling efforts to customer need and expectations.
However, machines merely help; humans take the final call. Human judgement can be brought into the equation by presenting the information with plain English explanations of the underlying reasons. This element of human judgement can make a huge difference in rooting out fraud without overwhelming a system. No one knows your business or customers better than you, and businesses can make better decisions when they know why a transaction has or has not been approved by the system. This means that banks will aim to ensure that they don’t stop legitimate parties from opening accounts, for example.
One of machine learning’s key features is its ability to detect patterns and recognise small deviations that occur which seem irrelevant to create an intricate profile. This functionality is perfect for supporting financial crime mitigation. For example, location data can be collected from a customer’s phone on an ongoing basis; looking beyond just a one-time snapshot of a mobile payment transaction. This machine learning approach would enable a profile of habitual behaviour for customers’ mobile payments activity. As a result, the profile is more accurate with alert detection, and false positives are reduced.
As well as increased profile accuracy, machine learning could also offer greater efficiency (and accuracy) when investigating possible fraudulent transactions. Many existing fraud systems have no automated decision workflow and rely mostly on manual review. This makes banks vulnerable to exploitation.
Adam Gable, product director – Financial Crime Mitigation at Temenos, highlights how his team is working with clients to combat fraud using machine learning. “Machine learning has been proven to reduce the number of false positives by five times,” he says. “This frees up time for investigation departments to deal with genuine problem transactions and avoids the need to inconvenience customers unnecessarily, supporting both external and internal fraud mitigation efficiencies. Sophisticated fraud modules are combing AI-based, self-learning algorithms and expert business rules to block suspicious transactions based on real-time behavioural analysis, allowing clients to identify and trap fraudulent and money laundering transactions as they occur. It uniquely builds customer profiles to detect and stop suspicious transactions that deviate from normal and expected behaviour.”
We are entering a new age of AI applications which are based on machine learning as the core technology. The current generation of AI systems offer tremendous benefits, but their effectiveness will be limited by the machine’s inability to explain its decisions and actions to users. Explainable AI, which is sometimes called XAI, will be essential for users to understand, trust, and manage this incoming generation of artificially intelligent partners.