The costs of financial crime (fincrime) are enormous for financial institutions. In addition, the operational costs of regulatory compliance and non-compliance penalties are onerous, reaching a pinnacle in 2020 to exceed $10 billion. The human toll of rampant fincrime on society is even more frightening.
Even in today’s more regulated environment, between 2% and 5% of global GDP (€715 billion and €1.87 trillion) is laundered annually, with the proceeds gained through activities such as political corruption, forced prostitution, drug dealing and illegal wildlife trafficking or used to fund terrorism.
Indeed, South Africa is under unwelcome international scrutiny due to laundering proceeds from state capture. Weak controls against money laundering landed the country on the Financial Action Task Force’s (FATF’s) grey list in early 2023, while money laundering cases against senior politicians damage public trust in the government and the financial system.
Manual systems are no longer good enough
Against the backdrop of tougher regulations and more sophisticated fincrime schemes, manual measures and legacy systems are no longer fit for purpose. Investigating vast data sets to flag suspicious activity should be automated, real risk should be escalated in real time, and historical trends in a user’s or business’s transactional behaviours should be identified.
While Anti-Money Laundering (AML) and Know Your Customer (KYC) checks are critical, financial services companies have identified resources and cost reduction as primary operational concerns. Reducing the costs of manual checks without compromising crime detection calls for more accurate data processing.
Automated systems mean that transaction monitoring can identify high-risk behaviours around the clock, conducting checks quicker, more consistently, and more accurately than a human. After all, crime never sleeps, and nor should financial services’ commitment to necessary AML protocols.
Today’s smarter transaction monitoring methods make practical use of Artificial Intelligence (AI). AI’s ability to process advanced data, automate, and detect anomalies removes the roadblocks traditionally associated with manual work, including false positives. A synergy of human expertise and AI technology answers inefficient compliance teams and processes.
AI’s fitting role in transaction monitoring
AI’s data analysis capabilities are invaluable to transaction monitoring. By identifying the context or characteristics of specific funds, predicting future account behaviour, and even tagging or indexing transaction data for visualisation purposes, analysts are far less likely to miss odd payments or build incorrect risk profiles.
AI and Machine Learning (ML) allow companies to move from fixed rules to a more complex rolling monitoring method, automatically combing through a customer’s spending patterns and historical behaviours over a long period. AI’s automation and predictive abilities are vital for reducing false positives and detecting anomalies.
False positive reduction
Reducing false positives with AI involves leveraging the results of previous alerts (resolved as either suspicious or not) to predict if any newly raised alerts are also likely to be suspicious or benign. This is typically formatted as a risk rating.
Suppose alerts are below a certain risk threshold. In that case, transactions have a very low likelihood of posing any risk, so the system can automatically close them. This helps a transaction monitoring compliance team manage their workload and only consider the highest-ranked risk alerts for further attention.
Anomaly detection
To discover transactions that could be identified as abnormal, AI processes a wave of historical transactional data to see if recent behaviour from an account (or groups of accounts) is behaving outside of what is expected from their past actions.
Through this process, behaviours that have been missed or unidentified can be flagged for their unusual appearance. The system can detect strange criminal typologies as they evolve. This helps to detect structured behaviours that launderers knowingly employ to evade detection in common rulesets.
Transparency and interpretability
The use of AI, with its immense data analysis capabilities, may hinder effective communication between humans and the system. Interpretability, crucial in machine learning, determines how understandable a model’s decisions are.
Transparency in decision-making is vital ethically, ensuring data interpretation is accessible to all. In AML, explaining transaction decisions is crucial, as regulatory scrutiny demands clarity to avoid fines. Algorithm selection can balance human interpretability with predictive accuracy in models, promoting responsible decision-making beyond blind reliance on AI capabilities.
Conclusion: redefined transaction monitoring
Despite fincrime’s ongoing threat, the regulatory reaction is encouraging. While the scrutiny and audit requirements from watchdogs may intensify to hold financial businesses to account, there are also encouraging signs in the uptake of modern AML systems.
In 2023, regulatory fines were reported to have indicated a positive movement in compliance reporting. Part of that comprises transaction monitoring, a task improved through AI automation and its advanced data-processing power.
AI is not about taking over the expertise of compliance officers and data analysts. As the enabler to streamline operational workflows for transaction monitoring using accurate data sets, it ensures high-value tasks retain focus.
Alongside trusted technology partners and AI experts, institutions can focus on what matters: strategically developing a product and process that can battle nefarious activity at the source.