Sunday, June 23, 2024

UnionBank uses AI to curb financial crimes

UnionBank uses AI to curb financial crimes

How do you feel about this story?



In the recently concluded “Next Generation of AI Transformation for Financial Crime Management in Asia” webinar, Union Bank of the Philippines (UnionBank) Head of Data Science Solutions II Josh Bosiños talked about the role of Artificial Intelligence (AI) bank solutions to curb financial crimes.

Mr. Bosiños highlighted that data is key in creating an AI solution to solve problems. “Focus first on the areas where you think you need to put more attention to, then once you have a clearer objective, we can now ask ourselves, so what data is available for us for this particular problem or challenge that we want to solve? Because in the absence of the data that we need, we can’t come up with any AI solution…,” he explained.

UnionBank has been developing AI solutions strategies in two main areas. First is detection of suspicious transactions alongside management of Anti-Money Laundering (AML) operations. Leveraging on the power of AI, the Bank’s investigators are assisted to look for patterns of suspicious transactions. This is on top of what UnionBank has already generated from its own alert system, thus, AI helps to avoid overwhelming the bank’s monitors.

Another strategy is using AI for resource management, using the data generated from the bank’s alert system. With warnings that are likely to be real, a true suspicious transaction will be initially investigated by the AI to ensure the reported alert.

The bank can also decide on a cut-off on the notices that will eventually lower the false positive alerts, a common challenge for a number of financial institutions. This specific AI program enables operational efficiency by reducing alerts to be manually investigated, thereby minimizing manpower resources dedicated to AML operations.

Mr. Bosiños likewise discussed UnionBank’s AI solution implementation for financial crime management on real time detection of credit card fraud and phishing.

Data Science experts designed the card frauds model through months of development. “The main challenge was how to link our available data with our full transactions database since this was an important part of the modeling process,” Bosiños shared.

Now, we have another reliable card fraud system capable of doing real time scoring. Given the volatility of the behavior of customers, the performance of this model is being continuously monitored and evaluated.

UnionBank has also established a similar model for detecting phishing incidents, which is also capable of scoring transactions in real time.