Financial institutions should embrace several sub-disciplines of AI in combatting fraud and financial crimes. These techniques will allow institutions to more effectively authenticate customers, improve customer experience, and reduce the cost of maintaining acceptable levels of fraud risk, particularly in digital channels.
Machine learning is a proven method that automates some of the supervised learning techniques in areas of fraud, with good training data on fraud events. We’re now seeing these approaches like decision trees, neural networks and GBM models being applied in anti-money laundering to predict “productive events.” Some of the advancements in linguistic analysis and contextual text analytics are proving helpful to automate tasks that have been historically performed manually. Any time you can reduce false positives by 50-70% with automated machine learning strategies, you’re freeing up precious human resources that can focus on more complex and subjective investigations.