Monitoring and investigating fraud is crucial, but time and people intensive, for banks. Now they are employing artificial intelligence as part of their weaponry, reports Joy Macknight.
Fighting fraud is an extremely laborious and costly exercise for banks and, historically, they have thrown bodies at the monitoring and investigation process. But this is no longer tenable as transactions move online, surge in volume and happen in near real time – whether payments or online applications. Fraudsters are also becoming more sophisticated.
Advanced artificial intelligence (AI) techniques – machine learning and deep learning – are making substantial headway in detecting and preventing fraud. Banks are now able to reduce headcount and increase productivity in compliance, due to significant reductions in false positives and improvements in detection rates. By leveraging AI tools, banks can review transactions – and all the relevant data associated with those transactions – significantly faster than a team of investigators could.
By deploying next-generation AI algorithms, banks can more accurately and intelligently identify and predict fraud by using event data and marrying it to real-world scenarios. Many are now looking to implement real-time analytics solutions that can prevent fraud without disturbing the customer, which is incredibly important in a world where user experience is paramount.
However, banks need to overcome two major challenges to take full advantage of AI’s benefits. First is to gather in one place the vast amounts of customer data needed to train the algorithms. Many banks are solving this problem by building data lakes or big data platforms. Second is being able to attract and retain staff with the right skills. As data scientists are expensive and in high demand, some banks are upskilling their investigation officers to decipher why a transaction was flagged by an AI algorithm.
Machine learning and deep learning algorithms find relationships among the data that might not be immediately observable. But in a regulated market such as banking, it is not possible to have a ‘black box’ situation where data is put in and a result generated without the institution being able to explain the decision process.
But even this problem can be solved by AI, and ‘explainable AI’ is an emerging research field exploring more explainable models and results for better human understanding.
