Most financial institutions are already on the path to using artificial intelligence in markets surveillance. NICE Actimize’s David Ackerman explores how this shift will impact compliance and the new skills compliance officers will need to cultivate.
As those tasked with financial market compliance will tell you, the vast majority of modern-day surveillance methods are predicated on analyzing previous malicious behaviors. By understanding what actions, events and consequences took place leading up to a manipulative incident, people and technology are taught to look for similar patterns. The glaring fissure in this approach, however, is the fact that you often don’t know what you don’t know.
Since current surveillance methods usually require a previous event to learn from, firms can be exposed to untold amounts of risk from “innovative” corruption that is either new or makes use of clever workarounds to defy pattern recognition. Modern regulation has attempted to address this gap by requiring financial institutions to record mammoth amounts of data particularly on communications and trades. Yet despite the fact that more data is being recorded now than ever before, market abuses like insider trading, spoofing and ramping continue to occur. What is the solution to this challenge? More and more firms believe the answer lies in applications that utilize artificial intelligence (AI).
The Path to AI Has Already Begun
Artificial Intelligence is no longer simply a point of discussion between financial institutions and regulators; it is a powerful compliance tool that has made important inroads in achieving regulatory compliance and streamlining operations. In a recent survey conducted by NICE Actimize, 89 percent of participants indicated they are already on the path to including AI in their compliance platforms. This is an encouraging trend from a legal and compliance perspective. For far too long, the financial services ecosystem has suffered from big data hubris, or the mistaken belief that more information is always better information.
Gradually, the industry recognized that big data does not necessarily translate to actionable information, and now a shift in strategy is occurring in real time to achieve that advantage. Breakthroughs in AI are allowing firms to focus less on capturing greater volumes of data and more on the story the data tells. With this new mentality, it is imperative to create clearly defined goals and a plan for a path forward that is different than past approaches to compliance.
Technology exists today that leverages machine learning (ML), a subset of AI that improves your ability to find information that you could not find before. While traditional surveillance methods look for a predefined scenario to repeat itself, ML allows for the discovery of new information by analyzing raw data for something outside of the norm. ML is a necessary component of an AI-enhanced compliance program, but ML is only once piece of a greater puzzle. Artificial Intelligence will take an anomaly and decide what to do with it. Far more than just simply if/than logic, AI asks how can this new information make the compliance process better. This activity can take many forms, such as alerting proper parties to the revelation, correlating the discovery with additional related information and creating a complete picture for a compliance officer to visualize or interpret.
Some applications are available in practice today; however, the level of sophistication and application vary widely. For example, 35 percent of respondents to the NICE Actimize survey noted that false positives remain a concern. Managing false positives is the bane of many a compliance executive’s existence, since even the most sophisticated surveillance programs require constant attention to keep pace with changing trading patterns and variable market conditions. Today, many firms use some form of supervised ML to improve the tuning of existing surveillance models. Compliance personnel will take the data returned by a ML algorithm and, in most cases, manually make adjustments to various thresholds of models in use (such as adjusting a look-back window or altering the dollar amount of transactions that triggers a particular alert).
The future will incorporate AI functionality into the tuning process, allowing for the compliance software to actively and automatically manage the end-to-end process. As self-tuning models become the norm, tuning as a function or workstream will become obsolete, thus freeing human capital for other endeavors.
Pushing a Square Peg into a Round Hole
Attempting to fit a transformative technology like AI into policies and procedures designed for traditional surveillance methods is arguably the single most common roadblock experienced by firms today. Imagine you own a horse and buggy. You plan your day accordingly, growing very comfortable and secure in your routine. Now imagine you buy a Tesla roadster. How realistic is it to expect the routine developed for the capabilities and limitations of a horse and buggy would translate to your updated method of transportation? It may sound like an extreme example, but when firms go from manual processes and legacy systems to AI technology, the comparison is not far off.
Take the voice surveillance requirements of swap dealers and other regulated users, for instance. A commonly used procedure requires compliance analysts to manually review phone calls, emails and chats. In practice, this consists of listening to hours of audio calls, reading every related email and scanning thousands of chats. The compliance officer summarizes the information and makes a determination as to the proper path forward. With the volumes of data described above, there is little hope most institutions can keep pace, and there is a high risk of integration point breakdown.
Firms increasingly look to AI when addressing these growing concerns. Machines are learning how to read as technology evolves. Natural language understanding (NLU) allows for automated data review, correlation and the summarization of the entire end-to-end conversation, complete with trade data. Incorporating NLU into a firm’s AI strategy eliminates the countless hours it takes to simply compile all this information in order to review it, and it allows compliance officers to investigate greater numbers of alerts in a more effective manner. Once NLU technology is active, the surrounding steps associated with data correlation, initial evaluation and summarization will need substantial changes. What formerly was a system associated with unearthing data can now be transformed into a validation or authentication routine. No matter what technological changes are adopted, it is imperative to remember: As the evolution of technology changes what compliance programs of capable of surveilling, so to must the policies and procedures put in place that govern their use.
Assemble a Team of 21st Century Compliance Officers
Although the overwhelming majority of participants indicated the inclusion of AI in their compliance process, nearly half of participants in the Actimize study are in the early stages of AI adoption. The use of new technology requires new thinking, and new thinking requires the prioritization of additional skills. Many firms are learning hard lessons as they trudge through the discovery and implementation process, but the learning curve will not end there. Future compliance teams will revolve around substantive legal knowledge, an understanding of financial instruments/marketplaces and an awareness of technology’s role in the surveillance process. In today’s landscape, few compliance executives have a mastery of all three skills.
For firms to fill the impending knowledge gap, they must encourage the development of next-generation technological fluency among their compliance officers and data management teams. This can be achieved in a variety of ways. Traditional methods include the attendance of conferences, tuition-reimbursement programs and the use of outside consultants who can help pave the way with less downtime. In a growing trend, many firms are partnering with trusted vendors to educate their staff on current and budding technological advancements. Regardless of what method is chosen, it is imperative that compliance teams develop the ability to understand, embrace and offer constructive feedback for AI integration.
Finally, it is important to note there is no substitute for human interaction. The fintech world is generations away from removing human interaction from the decision-making process entirely, so the investment in people is just as important as the investment in process. Artificial intelligence, machine learning, anomaly detection – these are tools. Like any other tool, they are only effective if you use them properly. Compliance personnel will likely not be removed in large numbers due to technological advancements, but instead be repurposed to maximize their efficacy.
Artificial intelligence and the innovations that accompany it will result in a level of operational efficiency long sought after by compliance programs around the globe. The resulting combination of upgraded technology, processes and people will undoubtedly strengthen the safety, soundness and integrity of our global markets for years to come.