Optimization of AML Screening Systems
Complidata is developing a platform with several AI modules which helps Financial Institutions detect financial crime more efficiently and effectively. Our platform consists of several “intelligent agent” that can automate or enhance different processes of financial institutions to better detect potential money laundering cases. Some of our customers call us the Robocop of financial services. One of the issues that we address is the problem with False Positive Alerts generated by screening systems. Legacy screening system will screen all transaction to know if these transactions are illicit so to detect potential money laundering or terrorism financing. These screening systems are generating alerts that need to be investigated by the financial institutions. Most of these alerts generated by the systems end up after investigations to be false positives, it is not abnormal to have around 90% false positive rate with current systems. We have devised a methodology based on machine learning that tries to predict if an alert will be a True Positive or a False Positive, true potential money laundering case or a genuine transaction. We are given alerts generated by the screening systems a new risk score, which allows us to rerank alerts on their potential of being a real money laundering case and therefore tell the financial instituions which alerts they should investigate to make sure to catch all of the true potential money laundering cases. The methodology also provides for reasons why our model has decided that these generated alerts are true or false cases, providing a narrative to the financial institutions on the reason why alerts can be deemed false positive or true postive. By using our methodology, we can reduce the workload of investigators with 40-60%, making sure that investigators can concentrate on the more serious cases, cover more areas where money laundering can happen, be more effective and efficient in their fight against financial crime.