Mind Your Language: Surveillance Systems Tackle Conduct Risk

For years, voice surveillance has been an integral tool for keeping tabs on traders’ activities. Whether for regulatory compliance or fraud detection, advanced analytical software has been deployed across all major banks to pick out bad apples from the market. 

Now, sophisticated voice technologies are letting banks go a step further to look at and analyze behaviors more commonly associated with conduct risk. Partly, this is being driven by external pressures on the financial services sectors, such as the UK’s Senior Managers Regime (SMR) from the Financial Conduct Authority (FCA) which first came into force at banks in 2016, as well as the global #MeToo campaign, which encourages victims of sexual abuse to speak up. 

Demand around the ability to study the deep complexity of human interaction and accurately alert banks to potential conduct risk scenarios is rising. Recently, Madrid-based surveillance provider Fonetic reported a 20 percent increase in requests from banks to address conduct risk by monitoring calls and chats between traders and other employees, compared with the previous year. 

“We have been developing new policies about profanity recently,” says a source at a tier-one bank. “In the past we also did [monitoring of] some non-authorized trading activities. That is our main field. Non-authorized could be from no proper language used, or about a client, or an activity in block trading, or market abuse, or whatever is not authorized by internal policy.”

Language Barriers

It might sound odd for a bank to be interested in monitoring employees using profanity. While swear words show lack of professionalism, does it really merit such a high-level scrutiny? 

“If you see the latest fraud misconduct, like [Libor, the London Interbank Offered Rate] or the foreign-exchange (FX) scandal, what we notice is the language around this non-authorized activity sometimes was full of profanity,” says the bank source. “There is a sort of language that is sort of bullish that sometimes is correlated. Not always, but sometimes it is correlated to this misconduct.”  

The scandal over the rigging of foreign exchange markets and reference rates resulted in hefty fines for major banks such as UBS, JPMorgan, Citigroup, Barclays, and Bank of America. One of the behaviors that was noted about the traders involved in the scandal, and published by regulators in all its lurid verbatim detail, was their use of swear words in private chat room conversations. 

“With the FX scandal, you notice when they speak together, they use bad words,” says the bank source. “They use slang words. I don’t know why, because they are educated people, but when they do sometimes it is sort of a bullish conversation. I am not saying it is always the case, but it is also another element that you want to consider once you are building up your policy.” 

Perhaps more destructive than the financial penalties, most of which were severe, but ultimately easily absorbed by the penalized institutions, was the reputational damage they incurred. It couldn’t have come at a worse time—following the financial crisis and global protests by another largely online-organized element in the form of the Occupy protests, the reputation of the financial services industry sunk to an all-time low. Evidence that bank personnel had deliberately fixed rates on which everyday citizens depended for their mortgage payments was akin to being kicked while they were down.

Now, reputational scandal threatens to rear its head again. While the brunt of the #MeToo campaign has been felt in entertainment, media and politics, the financial services industry has been no stranger to scandal in this region either, whether that’s accusations of sexual assault by staff at major banks, or through exposés such as the Financial Times sending undercover reporters to the Presidents Club dinners.

The use of communications-monitoring software to guard against harassment, bullying and assault—something the industry has long been accused of but rarely confronted—is in its infancy, the bank executive says. But it is being investigated.

“Definitely we see in the policy there are opportunities also to use it for this, and definitely this is something that the bank will have to look into in more detail. We notice that there are things happening on this side in the communications,” the source says. 

Abnormal Behavior

One of the top players in the surveillance space is Nice Actimize, which counts tier-one banks and other buy-side and sell-side firms among its clients. Steve LaGalbo is the director for communications surveillance at the firm. He explains that it was around the time of Dodd–Frank—the 2010 US banking rules that came into force after the Global Financial Crisis—that regulators started demanding that financial services firms to take more proactive measures to monitor voice conversations. 

“We created a solution that had the ability to connect to our Nice recording platform to analyze those phone calls, and to potentially uncover some risky conversations based on the conversation topics that were identified,” he says.

Over several years, Nice has evolved its product to be more of a complete surveillance platform, combining the electronic and voice communications surveillance capabilities into one product. “We are aggregating all the data from the trading activity, communications activity—including voice and other profile details—that helps us to understand what we call abnormal behaviors. We build profiles about the way people communicate, who they are communicating with, the types of things that we hear in the conversation, and we start to analyze those conversations to look for changes in those behaviors, or deviations from the norm.”

The idea is to look for anomalous individuals that are behaving abnormally, leveraging the data and analytics from Nice’s platform to give compliance teams a way to look at their environments in different ways to uncover conduct risk.

LaGalbo gives the example of Libor, monitoring the benchmark rate setting periods for the various currencies. In the old way of doing things, an analyst could just look at keywords or phrases that people might say that could suggest there is something suspicious going on. But what if the conversation is taken offline? What if there was no way of actually picking up what a trader says? 

“If we are looking at behaviors, and having all of this data in our solution, we could see that people involved in this activity always use certain chat communications,” says LaGalbo. “They always send a number of emails. They have a certain number of phone calls. There is a normal pattern of behavior for people to conduct business. Now if all of a sudden you have a trader who is not using their communication channels like everyone else—for example, they don’t use their phone at all, so during this critical time of business this particular individual is not using their phone and everybody else is—that is kind of suspicious behavior and could be a potential risk.” 

Nice uses transcription capabilities to convert everything that is spoken in the conversations to text. It also has analytics that highlight the concepts, the topics of discussion, and the sentiment—whether it is a positive or negative type of conversation. “If they want to look for conversations that have business context, they can very easily do that. If they want to look for conversations that might be suggesting somebody is harassing somebody based on some sexual misconduct language, then that is very well possible as well. Those types of analysis are done on phone conversations and can be highlighted as potential conduct risk.” 

While he says he has not had many inquiries on harassment, LaGalbo points out he doesn’t have 100 percent visibility as to how the customer is using the platform. “They could very well be leveraging our advanced contextual search capabilities on voice to be able to look for conversations that might include topics that suggest there is that sort of negative conduct happening,” he says.

Fit-and-Proper

The growth in the conduct risk space is attracting more players. Cloud9 Technologies, for instance, uses emerging technologies such as natural-language processing to power keyword spotting, which when paired with metadata such as who is on a call, how long it lasts, and similar information, can power a surveillance function.

“It also covers sentiment analysis, for example whether or not the tone of this person is one of anger or one of frustration,” says German Soto Sanchez, president at Cloud9.

Irisium, another surveillance provider partly owned by KRM22 and Cinnober, focuses primarily on market surveillance solutions to ensure that firms can comply with their regulatory obligations.  In 2019, it is launching new capabilities that will include communications surveillance, covering both electronic and voice communications. 

“Irisium has been watching very closely the development of the regulatory requirements on conduct risk specifically,” says Saeed Patel, director of product strategy at the firm. “The FCA in the UK is the first major regulator to move toward having completely new conduct rules associated with senior managers and actually having a new individual accountability regime in place.”

While the SMR was designed in the wake of the FX and rate-rigging scandals, regulators have definitively placed interpersonal conduct in the crosshairs as well. 

Recently there have been quite a few high-profile cases involving senior executives at banks who have been dismissed over sexual harassment, and the financial sector seems to be taking more notice. 

“[FCA supervision director] Megan Butler has confirmed that the new Senior Managers Regime will include allegations of sexual misconduct among the factors that look at whether or not someone is fit and proper to work in financial services,” says Jane Walshe, co-founder and CEO of regulatory intelligence firm Enforced, referring to comments made by Butler in May 2018 to the UK Parliament’s Women and Equalities Committee, in which she said the definition of “fit and proper” persons—a key definition under which someone or something is licensed to operate within the financial markets—would incorporate not just financial decision-making, but also culture. 

Yet at the same time as it has seen an uptake in requests from banks, Fonetic is also seeing a reduction in alerts. “Sexism is now less present in trader speech,” says COO Juan Diego Martin. “What could that mean? I don’t know. But it could mean that now traders know a real team is monitoring their conversation and they are starting to behave slightly better, let’s say.”   

The question is whether this has resulted in clear-cut cases that banks can pursue. The bank source, like others spoken to for this article, remains tight-lipped about the details, but says it has borne fruit.

“This is confidential information,” they say. “We do use it, and we do have some results. But I can’t talk you through internal cases.”