How effective is your fraud prevention software?
Most financial institutions use data mining (92.5%) or business rules management systems (BRMS) (65%) to address fraud, yet they say the effectiveness of these two solutions is only 28 percent and 18 percent respectively.
In our ongoing research series about AI use by financial institutions, AI Innovation Playbook: AI and Fraud reports on the experiences of more than 200 financial executives regarding the use of AI to prevent and detect fraud. The adoption rates are surprisingly low considering the potential for success, so we asked why.
Most agreed that artificial intelligence (AI) would be far superior (72%), but they have reservations. This final installment on the perceptions and realities of using AI in financial institutions reveals an interesting trend of perceived barriers: cost, complexity and transparency.
The firms don’t feel they understand the technology enough to gauge its effectiveness, and that includes the specialists charged with administering and evaluating FIs’ anti-fraud programs. Our research found that 60 percent of banks’ fraud specialists who use AI systems believe the technology is not transparent enough, and the same proportion view it as complicated and time-consuming.
Limitations can be overcome by adding Smart Agents, a powerful, distributed file system specifically designed to store knowledge and behaviors. This distributed architecture allows lightning speed response times (below 1 millisecond) and allows for unlimited scalability and resilience to disruption as it has no single point of failure.
FIs are keenly aware that the learning systems currently used to combat fraud are not up to the job. They express great interest in Smart Agents—AI-based systems that make real-time observations about interactions with human users— as the solutions would “know” account holders’ normal financial behaviors and could quickly spot unusual activity. More than two-thirds of surveyed fraud specialists view Smart Agents as ways to reduce manual review, a key priority for FIs in implementing learning system innovations. This may serve as a signpost for the way ahead: Specific applications like Smart Agents can turn AI from an abstract concept to a very real tool FIs can put to work for them.
The following are some of the key findings from our research:
- FIs see AI’s potential to more effectively fight fraud, but most don’t use it
- Just 5.5 percent of our sample banks employ “true” AI systems that can process and learn from large data sets and take personalized, case-specific actions
- Among those that deploy AI, only 45.5 percent use it as part of their fraud prevention efforts
- FIs believe AI’s real benefit is that it reduces manual review and exception processes
- 2 percent of FIs’ fraud specialists see reducing the need for manual review as a chief learning system benefit
- Surveyed FIs have concerns about AI’s complexity and transparency
- 60 percent of their fraud specialists feel the technology is not transparent enough, complicated and time-consuming
- The same specialists are more likely to fault data mining for its lack of adaptability (56.8 percent) and limited real-time functionality (48.6 percent)
- FIs express strong interest in using AI’s dynamic capabilities to improve fraud prevention
- 90 percent of respondents involved in fraud detection and analysis are at least “somewhat” interested in Smart Agents
- 7 percent believe the solution would help reduce manual review
- 9 percent believe Smart Agents would reduce payments fraud
Someone once said that the only thing in life that’s constant is change; that holds doubly true for bad actors in the fraud business. Only true AI with Smart Agents can stay one step ahead by constantly updating profiles and predicting fraud though anomaly identification. For FIs to stay ahead of the bad guys—and their competitors—adoption of true AI is key to consumer protection, profitability and growth.
Read the full report here to learn what else FIs told the analysts at PYMNTS.com, and what that means for fraud detection and prevention.