AI Made to Reduce False Positives
Damned if you do, damned if you don’t
Today, an institution can take two approaches to reduce false alarms. It can either lower risk thresholds to capture more suspicious activities, or it can tighten risk thresholds to lower the number of false positives. If an institution lowers its thresholds, the number of false positives increases; if it tightens its thresholds, the probability of missing fraud cases increases. With limited options, complex and changing obligations, and massive volumes of data to screen, the industry agrees that false positive rates are uncontrollable and compliance programs have become barely manageable.
My Celent report published yesterday, AI Made to Reduce False Positives: Detection Capabilities and Use Cases, explores how AI brings the promise of breaking the traditional tradeoff between false positive and false negative errors. Part 1 of this two-report series looks at the combination of AI technologies that will accelerate an institution's transaction monitoring systems and drastically reduce false positives, without missing true positives or compromising the institution's risk profile.
Celent predicts a rapid uptake of AI capabilities over the next couple of years to help institutions alleviate their compliance burden, including the reduction of false positives.
The report looks at 13 vendors' AI offerings for the reduction of false positives. What’s interesting is that each vendor has taken a slightly different approach, based on their POCs with banks. All vendors offer a level of advanced analysis and machine learning. Some focus on access to news content, watchlists, and data, where others focus on intelligent automation, robotic process automation, or more advanced segmentation analysis. Notably fewer vendors are developing natural language processing (NLP) and natural language generation (NLG) techniques. Celent believes the implementation of NLP and NLG tools are low-risk and low-cost, and suitable for the parsing, analysis, and construction of negative news content, regulatory filing narratives, and the generation of suspicious activity reports (SARs).
Vendors covered in the report include: @Arachnys, @Ayasdi, @Brighterion, @FICO, @IBM Financial Crimes Insight and Watson, @IBM Trusteer, @Intel Saffron, @LexisNexis Risk Solutions, @NICE Actimize, @Oracle, @Pelican, @Regulatory DataCorp (RDC), @SAS, and @ThetaRay.