The Risk Threshold Challenge

How AI Solves The Risk Dilemma

Traditionally banks have faced financial crime compliance dilemma: they can either lower risk thresholds to capture more suspicious activities, but also raise the number of false positives – or they can raise risk thresholds to lower the number of false positives – while also increasing the probability of missing true hits. This independent report by Celent looks at the unique role Artificial Intelligent (AI) technology can play in dramatically lowering risk thresholds and also deliver significant reductions in false positive rates.

Celent Findings

Pelican at the forefront FPR reduction

As this Celent report notes, 'Pelican continues to push the envelope in NLP and Machine learning’, leading the market in the application of AI in financial crime compliance and false positive rate reduction.
  • The financial services industry continues to struggle with out of control false positive numbers, with up to 99% of alerts erroneous
  • Europol estimates that only 10% of suspicious transaction reports (STRs) are further investigated after collection
  • Fortytwo Data estimates that large financial institutions spend in excess of £3 billion (US $4.6 billion) per year on AML screening alone

Download the Report

  • Celent reports that Pelican’s solutions have been battle-tested in high volume environments
  • Pelican solutions have been proven to reduce false positive rates by up to 72%
  • PelicanSecure has over 40 matching algorithms which generate detailed alert information, thereby reducing review times by up to 80%
  • Reinforced Learning reduces false positives - this stage analyses the behaviour of the operators processing alerts to understand and learn to continuously reduce the false positives.

Self-Learning benefits

Real-Time challenges

The Celent report highlights Pelican's unique AI and payments domain heritage and expertise. Noting the deployment of AI along the payments chain beyond correspondent banking to real time payments, where compliance poses significant challenges for financial institutions running batch processes.

Download the Report
  • PelicanSecure uses supervised Machine Learning to enhance the rules engine through self-calibration of results from analysts’ compliance review and reinforced learning to continuously incorporate the actions of analysts processing alerts
  • In addition, Pelican uses unsupervised Machine Learning with cluster and anomaly detection to identify new compliance techniques in real time
  • Pelican's financial crime compliance solutions uniquely contains in-built self-learning functionality which detects patterns of activity undertaken by compliance operators over time, particularly beneficial in real-time environments