From iPhone’s Siri to Amazon’s Alexa, from Google Maps to Netflix recommendations, and from customer service chatbots to the airplanes’ autopilot, artificial intelligence (AI) has seeped into our daily lives. It will soon impact every facet of our daily lives from driverless cars to smart homes.
Where AI is making a real difference is in catching criminals. Globally, United Nations Office on Drugs and Crime (UNODC) estimates that 2-5% of global GDP, or $800 billion to $2 trillion, is laundered annually. Money laundering is closely tied to other serious offenses such as terror financing, drug trade, human trafficking and corruption. This has prompted increased regulatory attention on financial institutions (FIs) which have struggled with their AML risks.
More than a third of FIs such as HSBC, MasterCard and PayPal have deployed AI to enhance their controls. Regionally, with several countries scheduled to undergo FATF mutual assessments, this offers a unique opportunity to the Middle-E ast FIs to implement AI based compliance solutions to monitor and report financial crime.
Companies like Ayasdi, BAE, Deloitte, IBM, Palantir, QuantaVerse and ThetaRay are some of the providers of AI solutions to the finanical services industry (FSI). In a recent Deloitte report on applications for the FSI, three main AI categories were identified: (1) Cognitive Engagement: AI which interacts with the users in natural language and acts on behalf of them (chatbots), (2) Cognitive Insights: AI which performs deep data analysis to provide insights (investigation tools), (3) Cognitive Automation: AI which replicates human actions and judgment with robotics and cognitive technologies (driverless cars). Some use cases of AI in fighting financial crime and sanction violations include:
Customer Risk Assessment: By combining and analyzing unstructured data from private and public sources (e.g. social media) AI helps build comprehensive profiles of customers and related parties to better assess their risks.
AML Transaction Monitoring System (TMS): Today, FIs are overwhelmed by the high number of ‘false positives’, i.e. false alerts, accounting for greater than 95% of all alerts identified by their TMS. AI based AML solutions can help reduce false positives significantly by learning from human supervisors.
Regulatory compliance: With regulations mandating Suspicious Activity Reports (SAR) filing within defined periods and fines issued for failure to do so, there is need to quickly investigate large amounts of data and the gap can be filled by AI developing baseline SARs.
Payment and Insurance Fraud: AI is moving to the frontlines by helping prevent fraud, where it is important to act before money is gone, through predictive analytics and identification of complex patterns in seemingly unrelated transactions.
Conduct risks (Insider Trading, Bribery and Corruption): AI can help analyze data across multiple channels and repositories including trades, calls, messages, etc. to detect suspicious staff behavior. Forensic accountants are using AI to detect inconsistencies and hidden relationships and uncover improper revenue recognition or weaknesses in business.
However, one must exercise caution before jumping onto the AI bandwagon. AI has an unproven track record. Misconfigured AI may have a high error rate. AI solutions are yet to be tested for the reliability generally expected from current solutions. One failure may lead to a systemic shock across the financial sector. AI solutions can also reinforce pre-existing biases, especially in the case of ‘human’ supervised learning. Thus, current AI algorithms cannot be trusted to make fully automated decisions.
As an AI program trains itself and evolves complex algorithms, it may not allow step-by-step explanation, thus offering lower transparency in decision making. AI solutions may infringe upon any customer privacy and data protection regulations. The EU General Data Protection Regulation prohibits fully automated decision-making that has a legal or similarly significant effect on an individual. As AI learns by training on live data, this raises important questions as to whether customers’ consent is required to allow use of their data for AI training.
For a Middle-East financial institution, AI has definite advantages over traditional rule-based analytics given the difficulty in finding skilled compliance officers. However, it must be combined with human instinct and local knowledge to ensure reliable and relevant outcomes.
In the end, AI is not just a tool to reduce ‘cost of compliance’. AI should be used to drive positive customer experience through faster approvals and lesser false declines while ensuring effective controls, thus, making it a bottom-line investment decision.
Nipun Srivastava is Director, Financial Advisory, Deloitte Middle East.