Artificial Intelligence and a New Era of Financial Services Compliance
BRG is home to renowned thought leaders and experts considered authorities in their fields of work. Our timely research and perspectives provide analysis and insights on the most important issues facing the industries and organizations we serve.
Prioritizing effectiveness, regulators are giving banks the green light to innovate on anti-money laundering efforts and are encouraging risk-based reviews.
Banks and financial service companies faced with escalating compliance costs are dealing with a particularly troublesome area—the fight against money laundering—where increasing volume and complexity of transactions mean that the stakes have never been higher.
Anti-money laundering (AML) and financial organizations’ broader efforts to stay on the right side of financial crime regulations require continued compliance investments—and can lead to potential fines if not carried out to a high standard. But there are human impacts as well. Human traffickers, drug runners, terrorists, cyber criminals and common con-artists operate in a world with increased political upheaval and have at their disposal an ever-expanding array of financial technology and new crypto currencies with which to launder their ill-gotten gains.
Both the stakes and the list of regulations have grown. Following the 2008 financial crisis, legislators and regulators unleashed a wave of new banking rules, including for AML compliance. Combined with the increasing number and complexity of transactions banks are processing each day, the growing regulatory burden has led to ballooning AML costs. Global spending on AML compliance was projected to hit $749 billion last year—an all-time high and an 18 percent jump over 2017.
The primary driver of those cost increases: headcount. The standard process for fraud prevention and detection remains highly manual, involving large teams of analysts and risk experts. Throwing more bodies at the problem hasn’t worked—it’s proven more costly than effective.
AML efforts through technology, not more manpower
At the Association of Certified AML Specialists (ACAMS) annual AML and Financial Crime Conference in September, it was noted that 90 percent of fraud alerts flagged in current transaction testing are false positives. As the number and complexity of transactions keeps increasing, bankers will have to face the fact that humans are simply not as efficient at detecting this type of behavior. These false positives result mostly from an overabundance of caution—a fear of missing something that might be real. And while understandable, and even reasonable, these conditions require large teams and more time to parse valid transactions from fraudulent ones.
The key to containing costs, reducing complexity and gaining better results may lie in today’s most advanced technologies: robotic process automation (RPA), machine learning and artificial intelligence (AI). Using AI-powered software for AML compliance can lessen a bank’s reliance on individuals, reduce headcount and free team members to work on higher-value, less repetitive tasks. It can also drastically cut the number of false-positive alerts and the risk of incurring steep fines for running afoul of AML rules.
Automating data collection with RPA bots can enhance “Know Your Customer” efforts, drawing data from many sources to build out more accurate customer profiles. Machine learning algorithms can be trained to parse varied data streams and formats to get all the data speaking the same language. AI can then use those inputs of automated and sorted data to model and predict classifiers and actions. The result: more accurate fraud alerts at lower cost with human capital freed up to focus on more thoughtful analysis (i.e., potential exceptions and the highest-risk customers and transactions).
Regulators want innovation in AML
In the past, bank leaders hesitated over investments in AML technology and innovative processes in part over fear that using new approaches might reveal past compliance mistakes or simply not work well enough to meet the expected compliance standard—either of which might expose the company to regulatory action.
Lately, it’s become clear that these concerns should not hold bankers back. Regulators are prioritizing effectiveness in AML compliance. A recent joint statement from the Board of Governors of the Federal Reserve System, Federal Deposit Insurance Corporation, Financial Crimes Enforcement Network (FinCEN), National Credit Union Administration, and Office of the Comptroller of the Currency says: “… The Agencies welcome these types of innovative approaches to further efforts to protect the financial system against illicit financial activity.” Further: “Pilot programs undertaken by banks, in conjunction with existing BSA/AML processes are an important means of testing and validating the effectiveness of innovative approaches. While the agencies may provide feedback… pilot programs that expose gaps in a BSA/AML compliance program will not necessarily result in supervisory action with respect to that program.”
Industry stakeholders are getting the message. At the ACAMS conference, speakers and attendees agreed that regulators expect innovation in AML and in most cases will not penalize innovative efforts. In some ways, the joint statement had the effect of officially sanctioning a period of experimentation, and it helped erase lingering fears that if investing in new AI methods revealed past mistakes, regulators would count those as strikes against the institution or impose fines.
Now, bankers and regulators can coalesce around a singular mission: more effective and efficient AML efforts.
For financial institutions, AI—broadly defined here as RPA and machine learning—holds tantalizing promise: nothing short of the opportunity to inverse the poor metrics of the current human-dependent process. But we remain far off from that moment. And it would be a mistake to view AI as a complete silver-bullet solution.
AML technology is promising, but still developing
Many banks start by running their normal AML processes and a parallel AI layer alongside and on top of the existing process, and then compare results. To justify the investment, the parallel trials must demonstrate that AI is at least as good if not better than its human counterparts—and that takes time. The models need to be trained. They need good data to ingest. And humans need to trust them. Those are big hurdles to standalone adoption, but recent trials look promising.
At this point, it’s almost a cliché, but still true: AI keeps getting better and better. It’s no longer controversial to embrace the technology, and everyone from the CFO to regulators understands the opportunities for those that can harness these technologies. Banks that can’t change their processes will become less efficient technically and financially, ultimately doomed to fall irreparably behind.
Compared to other industries, banking has lagged in technological adoption. At too many banks, mainframe computers are still coaxed into use from storage mezzanines. A truly disruptive force has yet to sweep the industry. But the race to develop AI for AML has started, and those that cross the finish line first will have the greatest advantage.