Fraud fighting's being harmed by too much manual labor
The number of digital payments being made across the globe is increasing dramatically. Unfortunately, this volume has been matched by an increase in the number of fraudulent incidents. In fact, fraud has reached the highest levels on record, affecting more organizations than ever. The scale of the problem was revealed in last year’s PWC Global Economic Crime and Fraud Survey. Nearly half (49%) of the 7,228 businesses across 123 territories that were interviewed reported that they had experienced fraud and economic crime over a two-year period. To deal with this change, more efficient risk-management tools and processes are evolving rapidly that leverage artificial intelligence (AI) and machine learning (ML).
Today, fraud management consists of several manual processes. Models and rules performance monitoring, fraud pattern discovery and fraud alert management are just a few. While these manual processes may be manageable at first, as the number of payment types and channels increase, it can rapidly become untenable to add more and more staff to manage and monitor all processes. With fraud on the rise, more rules and models also generally mean more alerts, which usually lead to higher incident rates. Combine the two and managing fraud can become very expensive, which is why efficient management processes are so important.
As ML assumes a larger role in detecting and preventing fraud, it is important to ensure that technology helps aid the end-to-end process — essentially utilizing machine learning as the autopilot. Just like when flying a plane, these systems sometimes require human intervention to check performance or respond to emergency situations by manually creating rules to stop a dangerous new fraud trend, for example.
Using "autopilot machine learning" to replace even one of the manual processes listed above can drastically reduce the amount of repetitive, expensive work a fraud analyst must do, allowing them to focus on more vital activities. Along with improving basic processes, the right level of manual intervention enables clear documentation of reasons for the machine's decisions. This also helps clarify future regulatory discussions and compliance. Maintaining a window into internal systems gives managers greater confidence in the machine's ability to manage processes and make the best decisions.
In fraud management, many of the same principles of autopilot flight apply. Between takeoff and landing, the autopilot can almost fly the plane completely without human intervention. The autopilot system relies on a series of sensors around the aircraft that retrieve information like speed, altitude and turbulence, which is then used to make the necessary changes. In the payments industry, as the technology used in fraud management becomes more and more refined, the fraud analyst will be able to release more and more control to the autopilot.
The autopilot example goes deeper. Pilots use autopilot to improve their own performance, increasing their work output, safety margins and productivity. After takeoff, some pilots apply extra care by flying the aircraft manually for a while to ensure that the aircraft is well-trimmed, and the flight path is stable before the autopilot is engaged. Again, the aviation analogy rings true for fraud prevention when adding a new customer, new payment type or going live with new channels. A period of human-powered hyper care is required to refine the fraud strategy before going on autopilot.
It's clear that in the future, certain elements of the wider fraud life cycle will be increasingly supported by AI and ML. Continuing with manual processes is not sustainable for the fraud risk management solutions of the future. The next generation of machine-powered fraud detection won’t be with new ML models, but in the way ML powers risk strategy management process automation, with autopilot ML.
ML for fraud management process automation is quickly becoming a necessity as the number of digital payments grows and fraudsters continue to innovate their way through the digital net. Smaller, more applied teams will be required in the future to work alongside the machine to create a system which is highly agile, performs well, and perhaps most important of all, is cost effective. Contrast this to today’s sprawling, inefficient manual fraud-management operations and it’s clear why ML is the way for fraud risk management's future.