AI-Powered System Aims to Streamline Credit Card Fraud Investigation
A new research framework is being tested to see if artificial intelligence can help solve one of banking's most time-consuming problems: sorting through thousands of potential credit card fraud cases.
The system, called Fraud Investigation Assistant (FIA), uses large language models to automate key steps in the investigation process. This includes gathering evidence and generating reports that a human reviewer can use.
The researchers behind FIA tested it on two established datasets and found that it improved the accuracy of fraud detection by 8% after investigating just 1,500 borderline cases.
FIA's developers say that it could help reduce the workload for human analysts, who currently face what they call 'alert fatigue' from dealing with an overwhelming number of potential fraud alerts.