AI Vault

Other “F” words: fraud in finance

Of all the “F” words that raise our collective blood pressure, none does so quite as effectively as “fraud.” It’s a crime that predates currency and, alongside society and technology, has only gotten more sophisticated with time. The increase of online banking and ecommerce has only exacerbated the opportunities for fraud. It has become a pervasive and costly problem with depth and complexity that makes it almost impossible to prevent. Fortunately, with artificial intelligence and its potent data intelligence capabilities, early detection of fraud is becoming a reality.

Needless to say, wherever there’s money, there’s usually a system to monitor its activity and check for fraud. These types of systems often track and gauge transaction activities, spending habits, or react to reported cases of fraud by those impacted. There’s already a certain degree of automation integrated in anti-fraud systems, but overall, manual human reviews remain the primary line of defense. This is a challenge because manual reviews of volumous financial data are error-prone, repetitive, and time consuming, not to mention costly due to wages and ongoing mandatory training of staff. In fact, according to a North American business survey, “manual review staff account for the largest single slice of [the] fraud management budget.”

Historically, fraud detection has been focused on discovering fraud rather than preventing it. This practice often results in higher rates of false positives or falsely-flagged transactions, which can erode or even destroy consumer trust. Merchants and banks lose out too on revenue as well as customers who, as a result of false flags, are hesitant to use their services or trust in their ability to protect data. In financial services, the technology already deployed to mitigate fraud leaves much to be desired. The traditional paradigm for fraud detection lies in customer history, where rule-based automated systems react to certain preset parameters. This can be a challenge because “false positives occur regularly with traditional rule-based anti-fraud measures, where the system flags anything that falls outside a given set of parameters.” Anti-fraud measures based purely on rules are devoid of human sensibility which is often essential to the understanding and mitigation of fraud. 

What sets machine learning systems apart is their ability to learn from massively diverse sets of data points, rather than merely focusing on customer history or a set of predetermined parameters. This way, sophisticated algorithmic models can build a more comprehensive profile for each customer and provide the right context to their purchasing history. Financial service juggernauts have already begun utilizing deep learning in their fraud detection systems, and by doing so, one company was able to cut down their number of false positives in half.

The implementation of AI in anti-fraud processes doesn’t entirely eliminate the need for the human touch. The human-AI feedback cycle necessitates a strong collaboration between people and machines. Many AI systems remain dependent on human input, especially in its early phases, and for integrating diversity into its models and understanding of data. Highly skilled analysts retain the ability to think like fraudsters and understand the complexity of human emotions that can affect how and why fraud is committed. In return, AI can help process vast, complex sets of data in a fraction of the time and do so around the clock.

These sorts of advantages are exactly what the financial industry has been looking for, especially in the wake of notorious financial fraud cases making headlines around the world. We’ve previously discussed how advanced AI can help streamline compliance processes, but at a business level, it provides even more. Leveraging the power of AI is a smart way to combat fraud and give peace of mind to concerned customers who are tired of hearing the “F” word.