AI solution powers radical shift in fraud prevention 

Fraud prevention rates in the global payments sector are set to be transformed by ACI Worldwide’s latest AI-powered technology, and using Azure has enabled the company to build its solution much faster and more effectively.

In fact, the solution was up and running within a year of the company’s data science team starting to investigate how to build it. “Using Azure definitely expedited us,” says Jimmy Hennessy, Director of Data Science and Software Engineering at ACI Worldwide. “We did not expect we could have an elastic, scalable service like this available to clients so quickly. Usually these things take years.”

ACI Worldwide built its patented new solution to address the complex and ever-evolving landscape of fraudulent transactions around the world, radically reducing the huge financial losses banks and retailers suffer because of fraud.

“Our goal was to be able to have a fraud detection model decide for itself when it needed to change to keep its performance at the same level as it was previously,” explains Hennessy. “Over time, fraud behaviours change and fraudsters do different things, so the historical data becomes less relevant and the model degrades.”

AI constantly learns how to beat fraud

His Limerick-based team, the core data science group for the global company, developed the machine learning technology, which all runs on Azure. It recently won Best Application of AI in a Large Enterprise at the Irish AI Awards. Based on a concept called incremental learning, the AI behind it not only keeps itself up-to-date, it actually improves itself, which the team had not expected it to do.

To explain incremental learning, Hennessy uses the analogy of someone learning to play piano. If that person learns how to play five pieces really well, they can go on and learn a sixth piece for piano without having to revisit the first five pieces.

From a client perspective, it is like a silent success,” he explains. “They enjoy less fraud, no disruption, and fewer chargebacks and false positives, but they don’t have to do anything to achieve that as it’s all automated.

While the AI can identify and adapt to new types of fraud, it also operates within set thresholds, so clients don’t suddenly have to contend with sudden, significant changes made without human intervention.

Achieving success with Microsoft expertise

ACI Worldwide is a global software company specialising in real-time payment and fraud detection solutions for financial services. While the team started working on ways to put incremental learning into practice in ACI Worldwide’s own production environment, the company also kicked off its partnership with Microsoft and began a five-year plan to move from on-premise data centres to Azure.

At that point, the data science team changed its approach architecturally and brought an Azure cloud architect on board, says Hennessy. He adds that the team also started engaging with Microsoft as part of its production work, and describes working with Microsoft subject matter experts as an iterative, collaborative process.

Our colleagues at Microsoft threw lots of ideas at us as we developed this. It has been great having a subject matter expert at the other end of a call when we run into problems, which might not even be anything to do with Azure. Sometimes, what would have been a day-long exercise of bug fixing for us, turns into a 20-minute call with them, and it’s sorted. So it has been a really healthy, positive engagement for us.

He adds that many of the data science team members would not have used Azure before, but have embraced it with this project, with some getting fully certified.

“Data scientists are an inquisitive breed,” he says, “and it really does generate a morale boost to investigate something new and build it in a relatively short period of time. Overall, this has been a truly energetic, passion-driven project.”