The future of communication surveillance: moving beyond lexicons

Detecting misconduct is complex. Communication surveillance continues to be one of the most complicated and challenging facets of regulatory compliance.

Communication surveillance involves the collection, monitoring, review and preservation of information that has been communicated across various internal and external channels within a regulated organisation. In highly regulated industries, compliance teams are often legally obligated to monitor their employees to ensure they don’t violate laws around insider trading, market abuse, money laundering, or any of dozens of forms of potential misconduct. Technology to monitor communication has become core to streamlining these once manual processes, but the most common solutions are often ineffective.

Lexicons are just the beginning

Technology that is used for surveillance traditionally uses lexicon triggered alerts to define the set of communications that require a review by a compliance analyst. Lexicons are words and phrases that are pre-determined to help reviewers seek out specific violations. For example, the word “guarantee” might indicate a registered representative making an inappropriate promise, or the word “confidential” might reveal insider dealing. Many firms use lexicons in their surveillance strategy because it’s simple and easy to defend.

While monitoring lexicons can be simple, there are challenges including high alert volume, false positives, alerts on irrelevant content such as disclaimers, and duplicative content. Additionally, a lexicon-only approach is point-in-time, so evolving behaviours could be missed. Metadata and advanced technologies can make a world of difference as they help to create better rules for a more accurate and efficient process.

Metadata fills in the gaps

While monitoring for lexicons is a great start for surveillance and compliance teams, applying metadata is crucial. Metadata is a set of data that describes and gives information about other data (i.e., context and background information). In surveillance, it’s context around a communication: who the participants are (such as their role, department or geography), the file type, the time of day the communication was sent, whether a message was inbound or outbound, and even the language spoken during the communication exchange.

To determine if collusion is occurring, helpful metadata might show how many participants were involved in the conversation, as collusive behaviour most often happens in small groups. Rules should only apply to the populations that are at risk for that specific behaviour, so for this instance, rules around the sharing of inside information should be targeting individuals with potential access to inside information. Another example of a red flag is changing venues of communication. Intentionally moving an inappropriate conversation away from a recorded channel is an indicator of risk. A client may say to a trader, “Let’s chat about this over drinks.” Fortunately, that conversation started in a recorded channel, which sets off an alert for further investigation. Getting more granular with metadata filtering will make the surveillance process more effective and efficient.

When lexicons and metadata aren’t enough

Advanced technology such as artificial intelligence (AI), machine learning (ML) and pre-trained models can better detect misconduct and pinpoint the types of risk that a business cares about. AI and ML should work alongside metadata filtering and lexicon alerting to remove irrelevant data and classify communications. In pinpointing risk, AI and ML can help to:

• Go beyond individual lexicons to detect misconduct
• Detect irrelevant and duplicative data to significantly reduce false positives
• Identify trends and patterns across the language used within communications, which are impossible to pick up by a group of reviewers
• Automatically learn over time by taking input from the team’s review of prior alerts
• Adapt quickly to changing language to identify phrases you didn’t know you needed to look for
• Build a new model or adjust an existing one, which only requires past examples of a behaviour

There is often an assumption that massive adoption of AI will automate the job of a human reviewer. That’s not true. AI and ML supplement compliance review teams and help them focus on more relevant communications and remove obvious junk from their review queue. The technology can transform the role from reviewing hundreds of alerts per hour to one that is more investigatory in nature. Financial institutions will benefit from adopting these technologies as they can eliminate some of the manual and monotonous work so employees can focus on the more meaningful aspects of their roles, such as stopping insider trading or collusion before it happens.

Take the next step

Surveillance technology that incorporates metadata and AI is already necessary to meet growing data volumes and an evolving regulatory landscape.When a company re-evaluates its surveillance strategy as regulations evolve, it’s crucial that the industry moves beyond lexicons and shifts to an AI-first platform to be the most effective.

If you’re interested in building an AI-based surveillance strategy to better identify risk, remove duplicative data and reduce false-positive alerts, check out Relativity Trace ebook, 3 Steps to Building an AI-based Surveillance Strategy.

by Jordan Domash, General Manager of Relativity Trace, Relativity.