More banks than ever are considering and adopting real-time payments, particularly with the recent launch of FedNow. One of the first questions they face is: “What about fraud?” Banks have never been more pressured to monitor, evaluate, and approve payments in real time, flagging fraudulent behavior (and also resolving insufficient funds exceptions) almost instantaneously.
Banks fully grasp the powerful imperative to adopt real-time payments: They want to enable convenient on-demand services that attract an increasingly savvy customer base. As banks examine how their traditional fraud prevention methods would facilitate real-time payments, they realize that their legacy systems are simply not up to the task. Fraud risks and technology barriers around instant settlement are leading some banks to be cautious about real-time payments adoption.
To de-risk real-time payments, banks need a technology-driven approach to fraud prevention that is reliable, cost-effective, and fast. On this front, artificial Intelligence can be their most powerful lever. By leveraging machine learning algorithms, banks can analyze real-time transactions and swiftly identify irregular patterns of behavior that could signal fraud or money laundering. Further, as they unlock the power of AI for fraud prevention, they will soon find themselves leveraging AI to add value in other payments-related use cases, including use cases that drive new revenue.
The Challenges of Real-time Payments
Amid all the benefits that real-time payments will bring to their customers, banks are understandably concerned about compromised payment screening.
Many of the concerns around payment screening are wrapped up in the larger technical challenges of supporting real-time payments. Integrating real-time payment screening into existing infrastructure can be complex, time-consuming, and fundamentally un-scalable. At the same time, migrating away from legacy data systems can be costly. Like most emerging digital capabilities, real-time payments journeys will be more seamless if they are embedded in larger digital transformations that take advantage of cloud-based solutions.
Also, any real-time payment screening solution – AI-driven or not – will need to consider the role that human experts play in the process. Human experts will continue to define the initial rules and parameters for fraud detection systems. Traditionally, they have played numerous crucial roles in determining the legitimacy of a flagged transaction: reviewing additional data, fetching customer records, contacting customers, and making judgment calls. In a real-time payments’ universe, much of that workflow can be safely automated, but complex edge cases will require periodic human intervention.
When transactions settle instantaneously, banks are taking a significant risk if they do not simultaneously enable instant fraud checks. With its continuous transaction monitoring capabilities, AI will help financial institutions flag potential fraud and stay on the right side of regulators as they ramp up their real-time payment volumes.
AI models that have been finetuned using previous fraud instances and suspicious activity reports can make predictions about fraudulent transactions by analyzing similarities between past instances and current transaction data. These models can detect patterns in customer behavior (such as cash flow and payment patterns), form criteria to identify potential fraud or money laundering, and raise red flags in the rare cases where transactions need to be quickly reviewed by bank employees.
Increasingly, AI will also support payment screening based on behavioral biometrics. Behavioral biometrics distinguish between legitimate users and cybercriminals by analyzing an individual's unique behavior and creating a profile based on that unique behavior that can include variables like typing speed, mouse movement, and online browsing patterns.
As their AI capabilities grow, banks can also leverage deep learning models to process complex unstructured data, such as images and text, further enriching the landscape of available datapoints.
An AI-driven approach will also be indispensable in preventing money laundering and upholding regulatory compliance for real-time payments. For example, banks will be able to further optimize fraud detection through AI-enabled analysis of payment beneficiary networks. According to McKinsey data, GenAI-driven approaches to fraud detection will also bring an efficiency boost, increasing productivity in fraud detection by 30 to 50 percent and reducing costs accordingly.
The Art of the Possible: AI Beyond Fraud Prevention
As banks enable real-time payments, fraud prevention will be the most critical application of AI. But they will also discover other use cases for AI in the payments journey.
For instance, AI-powered chatbots powered with insights derived from payments data can offer quick and precise assistance in navigating payment-related queries. This will enhance the user experience and alleviate the burden on customer support teams.
While mitigating payments fraud and providing new operational efficiencies, AI will also open new revenue opportunities for the bank, such as cross-selling convenient product offerings tailored to a customer’s needs.
AI will soon be a driving force that both facilitates the expansion of real-time payments and extends numerous payments-related opportunities to forward-thinking banks. Real-time payments are accelerating across the globe: FedNow in the US, UPI in India, FAST in Singapore. Banks, regulators, and customers all want the same thing: fast payment solutions that are secure, efficient, and personalized. For banks seeking to seize the full advantages of real-time payments, AI provides the path forward.