Editor’s Note: This week, we’re diving into how AI helps compliance teams keep up with changing regulatory priorities by reducing alerts so teams can focus on real risk. Read this two-part series to learn how technology helps by eliminating duplicative text, analyzing and contextualizing content, and transcribing multilingual communications.
A procession of misconduct scandals and controversies have eroded public faith in businesses in recent years. Regulators are increasing scrutiny of corporate behavior as a result, and the punishments for those who fail to meet expectations are dizzying. In 2021, the estimated cost of financial crime compliance for global financial institutions was projected to be around $214 billion.
Monitoring for compliance is one of the most important operations in any regulated company, but the job has become much harder by virtue of the same tools that allowed many companies to stay afloat during the pandemic: digital communications.
Digital chat and video functions play a vital role in connecting employees, but as these channels grow in number and complexity so does the challenge to compliance teams who must wrap the channels into their surveillance coverage. It’s estimated that 1.4 million video/voice calls are made every minute worldwide.
Misconduct flourishes where gaps in monitoring exist, and no team of human reviewers could hope to find every piece of meaningful content within the exponential amounts of communications produced each day. The increasing digitization of communication and workflow processes means more data, and more alerts, reviews, and occasions where a reviewer doesn’t have the full context of a conversation.
Forward-thinking teams have turned to artificial intelligence (AI) to help. AI can pinpoint misconduct inside avalanches of text, images, video, audio, and many other formats when trained to spot scenarios and behaviors.
Behavioral surveillance of communications enables profiles of conduct to be built over time, allowing compliance teams to identify more sophisticated and subtle indicators of misconduct. These deep-learning algorithms can uncover patterns invisible to human eyes and flag potential wrongdoing before it crystallizes into a significant breach.
Businesses that apply these techniques and tools can better protect their organization from misconduct risk, while reducing false positives by more than 90 percent compared to legacy systems.
Read on for a framework of how to get started with this approach.
Tip #1: Leverage a Single, Modern Platform
A single communication platform inside an organization is the key to optimal surveillance.
Funneling both structured and unstructured data into one place allows for a more complete view of an employee beyond the words they type or say. From here, artificial intelligence and natural language processing (NLP) algorithms can identify patterns of suspected or actual misconduct through deviations in behavior as evidenced in their communications.
This is a significant step toward shoring up holes in surveillance coverage that may have developed in recent years as more chat channels, such as mobile apps or video conferencing solutions, appear.
Unstructured data can account for at least 80 percent of workforce data. Unless tamed with the right system and analytics, this data can easily swamp any compliance team. Communications can have multiple authors, and words can be shorthand, slang, or substituted with emojis or images—all making risk harder to find. Surveillance has to cover chat platforms like Bloomberg, Reuters, and Symphony; enterprise chat systems Slack, Teams, and Skype; video conferencing tools such as Zoom and WebEx; and also file sharing and collaboration applications like OneDrive, Box, and SharePoint. All this comes on top of the standard surveillance of phone calls, text, email, et cetera.
Tip #2: Focus on Getting Rid of the Noise
Smarter tools can give you control over what is removed and what stays for analysis, allowing you to identify only suspicious scenarios. Non-authored content accounts for a majority of false positives, therefore AI that removes spam, newsletters, reports, and the like, greatly reduces alert volume—which is unlikely to include true risk.
Email threading, along with text cleansing, can further reduce document counts and alert volumes by up to 60 percent.
An email thread is a single email conversation that starts with an original email and includes all of the subsequent replies and forwards pertaining to that original email. If a surveillance tool alerts on content in the first email of a 30-email chain, that same content will trigger an alert in the following 29 emails. Offshoot conversations, signatures, disclaimers, forwarding, repeated data would all reappear, confusing the picture. This makes it harder for compliance staff to investigate, resulting in them poring over copies of the same data.
Email threading groups all emails in a single conversation thread and strips out the unique content. AI can determine if emails belong to the same thread, and if identical, will discard them for the purposes of monitoring. Having filtered out the junk, the system will then present the relevant information for review. This drastically reduces the number of alerts, preventing repetitive work while ensuring all the content is monitored.
The algorithms also scan attachments in emails. This is important because email signatures, server stamps, confidentiality footers, and other repeated content increase false positive alerting and can overload legacy systems. References to different types of misconduct in confidentiality footers are notorious for triggering false positives.
Tip #3: Implement Transparent, AI-powered Communication Surveillance Features
Many solutions on the market include email threading but do not allow you to control what is removed. That’s why black box surveillance solutions are discouraged by regulators. Alternately, being able to see what was cleared and why allows compliance teams to retain a flexible, defensible system of review.
A best-in-class solution should keep up with changing regulatory environments, while providing a transparent, defensible reviewing experience for compliance teams. By using AI-powered text cleansing, surveillance professionals can utilize email threading to get rid of irrelevant content and focus on real risk.
Relativity Trace goes above and beyond—in one year alone, Relativity Trace analyzed more than 2 million gigabytes, and our communication gathering tools are constantly expanding to intake data from a wider source of communication avenues. Our dedicated AI team also helps tailor the solution to the individual needs of each customer, understanding that there is no one silver bullet to eliminating every compliance challenge.
Stay tuned for the next installment to find out how AI helps pinpoint risk and provides context for reviewers.