At one of the world’s largest financial services companies, systems were monitored constantly using an engine that detected potentially fraudulent activity and determined which automated action should be executed. However, teams still needed to manually route events. As attacks continued to grow more frequent and sophisticated, the number of events that had to be inspected continued to expand, and the number of false alarms also continued to increase. These issues placed a growing strain on staff and led to increased cost and exposure.
Recognizing the potential to employ AI to address the fraud challenge, the company’s data scientists developed a composite fraud detection model by using a combination of existing rules-based and new machine learning (ML) models. However, this and other models were sitting in the labs, as it was very complicated and risky to push them into production without a proper ModelOps system.
