Focused on machine learning to fight fraud, this is the first article of a two-part series on the deployment of artificial intelligence in banking and questions surrounding the ethical use of data.
“It’s really important that we have more intelligence than artificial.” That was Apple co-founder Steve Wozniak’s thought as he discussed artificial intelligence (AI) at Money20/20. My takeaway was that regardless of how “smart” machines become, human intervention will always be required. When we examine AI in banking, that’s a powerful point.
Increased complexity surrounding data
AI in payments has grown in just one year from machine learning that fights fraud to encompassing every portion of the payment ecosystem. Such rapid diffusion of AI usage begs the question: How will the industry ensure that individuals’ data is used in a fair and responsible manner? Machines won’t be able to make those decisions without human intervention.
Trust is something financial institutions must maintain. FIS’ annual PACE study has repeatedly shown that bank customers consider their security the No. 1 attribute when it comes to trusting their bank and the No. 1 feature they expect their banks to deliver. However, with some banks fending off as many as one million cyber-attacks a day, the fight to maintain consumers’ security has become relentless.
Holders of data, including financial institutions, are under increased scrutiny due to the enormous jump in the number of data breaches. In 2016, the number of reported data breaches increased by 40 percent and more than nine billion data records have been lost or stolen in the past four years alone.
Machine learning is critical for delivering security and convenience
A recent Aite study, with senior fraud and data analytics executives, found that 68 percent of respondents place ‘very high’ priority on investments in machine-learning analytics for fraud mitigation.
FIS already leverages data from issuers to enable machine learning in the fight against fraud. Machines “learn” individuals’ patterns of behaviors – what they buy, where they shop, how much they spend – and, over time, become increasingly precise in flagging fraudulent transactions. Consumer controls that send alerts when transactions are flagged for potential fraud further help to enable cardholders to verify a transaction and help stop fraud in its tracks.
As a result of machine learning, the number of false positives – transactions mistakenly declined at the point of sale – drops, and fewer consumers experience the embarrassment of having their card rejected. That’s important because, as Javelin reported in a 2015 study, 15 percent of cardholders had at least one legitimate transaction declined in the past year, and 39 percent of those who did abandon their card completely afterward. This shows why mitigating false positives have a significant impact on interchange revenue and interest income for issuers.
Financial institutions no longer have to trade heightened security for customer convenience when making payments. As machine learning applications increasingly become “smarter,” it will allow consumers to enjoy both, without any compromise.
Click here to read part two and learn how to fight fraud in other areas of the payments ecosystem with artificial intelligence.