By Jordan Domash, General Manager, Relativity Trace.
Over the last 10 years, the term artificial intelligence (AI) has turned into a dinner table conversation topic despite the fact that most people cannot accurately define what AI is even in simple terms or truly understand all the possibilities of the technology. AI doesn’t need to be abstract or mysterious. While it has enabled humans to build self-driving cars and computers to recognize speech, it can also improve the efficacy of repetitive business processes that we encounter every day. I believe that over the next 5 years, we’ll see a massive transformation in regulatory compliance as teams embrace AI to become more efficient and to better detect misconduct.
For our purposes, let’s define AI as a computer system that supplements human capabilities by identifying patterns in data, learning from experiences and making statistically based decisions. Compliance officers use a variety of systems to review alerts of potential misconduct every day. In communication surveillance, these alerts indicate a risk of market abuse or market manipulation as detected from an employees’ email, chat, or audio communications. Compliance officers review these alerts and then make a decision – either it’s a true positive (needs to be escalated) or it’s a false positive (should be closed). This is an application that is ripe for a computer system to supplement human review.
Pointing back to our AI definition, compliance officers provide training for the algorithms by labelling each alert as “true positive” or “false positive.” The system will identify patterns and as compliance officers continue to give feedback on the AI model (through labelling good and bad alerts), the system will learn from experience ultimately making statistically based decisions on what should be an alert in the future. Over time, improved AI models will allow compliance officers to focus less on reviewing junk and obvious low-risk alerts and more on meaningful investigatory work to understand the context behind potentially high-risk employee activity.
A brief history of AI in legal and regulatory services
While AI has become more commonplace in other financial applications like high-frequency trading, compliance has maintained a level of skepticism. As the leader in e-discovery, Relativity has already seen an industry transformed by the adoption of AI technologies over the last 10 years. We released our initial analytics algorithms in 2008 to categorize documents and minimize time spent in the review process. At that time, we regularly encountered lawyers who pushed back on the defensibility of the new and evolving technology. But as lawyers began to adopt analytics into their processes, we saw an explosion in efficiency. The number of users in Relativity has grown 13x over the last 10 years while the amount of data has grown 121x. Each user can sift through exponentially more data now than they could before we introduced AI. Now, using AI to supplement an e-discovery review process is commonplace. In 2019 alone, lawyers used Relativity Analytics on more than two petabytes of data. That’s the equivalent of reading the dictionary more than 650 million times. And there is a suite of court cases such as Rio Tinto Plc vs. Vale S.A. or Pyrrho Investments Ltd. vs. MWB Property Ltd. that can be refenced to explain the defensibility of this technology.
Many lawyers who hesitated to use AI for many years have recently become more comfortable with culling and organizing evidentiary documents using the technology– it’s turning into the new normal. But this shift did not happen overnight; it took close to a decade to reach a significant level of comfort using digital analytics for it to become the standard.
Taking compliance to the next level with AI
Compliance, however, is a different story in its uptake of the technology. We expect that to change, and that leveraging AI will become a baseline when it comes to managing regulatory risk.
Today’s standard in monitoring technology is using simple lexicons (search terms), which as you might expect, delivers a massive number of false positives. There’s so much more that can be done today with AI to streamline this work. AI can help remove obviously irrelevant content like spam and duplicative content like repetitive emails in the same email thread (made up of forwards and replies). AI can help pinpoint risk through fine-tuned machine learning models that can identify collusion, insider trading or other types of misconduct. It can also make audio and image files searchable to understand and detect any hidden nefarious behavior. All of these tools aid compliance professionals in identifying market abuse and other risky behaviors. Compliance teams need to grow more comfortable with AI to keep up with the explosion of data around them or risk falling behind.
And while increasing efficiency in communication surveillance is important, it’s not just about cutting costs. Leveraging technology to supplement human searching can also reduce risk by identifying themes that may have slipped past a reviewer or not been caught by human-created search terms. AI offers capabilities to transform an organization’s risk mitigation strategy.
With AI, teams can be more efficient and reduce risk at the same time. We’ve seen it work before.
What does this all really mean?
Implementing a technology alone is not always enough – there is not just one algorithm, behavioral analytics tool or machine learning model that will solve surveillance challenges. It takes a suite of highly configurable capabilities that, when used together, will greatly reduce false positives and pinpoint organizational risk for compliance teams struggling to keep up with growing alert volumes.
Over the last few months, the world has experienced volatile financial markets, which has led to an increase in alerts that require review; now, there’s an even greater need to layer on this type of technology to add leverage to compliance teams. With this comes increased risk among traders and employees at those firms. A remote working environment provides people the opportunity to use unapproved communication channels (i.e. personal devices), which makes it more difficult to capture all their communications, in turn supporting riskier behavior.
With the proper AI strategy and a platform that offers an extensive suite of AI capabilities, individual surveillance team members are better equipped to remove irrelevant content, receive alerts on the riskiest behavior and unearth the truth. If organizations want to increase efficiency, reduce review time and quickly pinpoint misconduct before it escalates, now is the time to integrate AI into compliance and communication surveillance operations across the board.