In the wake of the historic fines handed down for manipulation of foreign exchange markets, banks are seeking solutions to prevent such collusive behaviour happening again. The challenge for financial institutions is finding ways to monitor the tens, if not hundreds, of millions of pieces of email, instant messages and phone transcriptions – a daunting task, to say the least.
"Individuals are engaged in multiple conversations simultaneously. They potentially write hundreds of emails a day, engage in numerous chat sessions and use electronic messaging services. Across a global banking institution, the sheer volume of these electronic communications becomes a surveillance nightmare without the right tools," says Rob Metcalf, president of Tennessee-based technology firm Digital Reasoning.
"One approach would be to hire a lot of people to read that stuff, but that doesn't work because of the scale of the data," he says.
This is where Digital Reasoning's Synthesys comes in. The platform takes a human-like approach to intelligence. As it learns, adapts and becomes smarter over time, the technology looks at communication in different ways in order to interpret the meaning of the data it processes, and it uses reasoning and experience to reveal any potential risks and threats.
"There are a number of techniques that absolutely would have identified examples of the communications that ended up being problematic," says Metcalf. "What we have done is to leverage some pretty interesting science and maths, figuring out ways to reduce communication to a machine-intelligible format, where we can train or teach a system to find risks and make predictions."
Since the communications involved in the collusion are in industry shorthand and sometimes coded in terms that are unintelligible to the uninitiated, the technology must learn to identify the relevance of the data it analyses.
"We understand that you may not care about a cable if it is in the context of what I'm going to pick up at an electronics store, but you will care when it's being discussed about a currency pair during an unusually volatile time of the trading day. It's that level of understanding that enables the computer to sift through a lot of otherwise noisy and unhelpful information," says Metcalf.
Once the platform detects a potentially problematic electronic message, it notifies the compliance manager and highlights what is being talked about, who is involved and what the concerns are. It is then the role of the manager to speak with the employee(s) in question, and judge whether the concerns are warranted and if further action is needed.
"We are not trying to replace people. We are taking the best of what humans can do and enabling them to focus their time on doing exactly that. The software is not taking on a prosecutorial role; it is surfacing communications with certain patterns that are particularly problematic," says Metcalf.
This article was first published in FX Week, and is used here, in its entirety, with permission from FX Week.