Lee Garf, general manager, financial markets compliance for NICE Actimize, takes questions about the new technologies that will help firms fight financial crime. The company won the Best Financial Crime Prevention Technology award as part of the FTF News Technology Innovation Awards for 2018.
(Advanced technologies such as behavioral analytics, machine learning, artificial intelligence (AI), natural language processing and cutting-edge behavioral analytics are going to be game-changers in the ever-expanding challenge to detect and uproot financial crime, says Lee Garf, general manager, financial markets compliance for NICE Actimize. He took time out of his very busy day to answer our questions in the wake of NICE Actimize winning the Best Financial Crime Prevention Technology award as part of the FTF News Technology Innovation Awards for 2018.)
Lee Garf, general manager, financial markets compliance, NICE Actimize
Q: What would you say was NICE Actimize’s biggest achievement in securities operations 2017?
A: NICE Actimize’s launch of its newest surveillance solutions using behavioral analytics and machine learning was one of our most significant accomplishments in the past year, one that continues to bring strong results to our customer base on both the buy side and the sell side.
One financial institution customer recently applied these solutions to uncover insider trading and employee theft of confidential information by identifying changes or anomalies in an individual’s behavior. Our behavioral analytics quite efficiently and effectively led them to the source of the problem.
Q: How would you characterize the current state of anti-financial crime regulation for financial services firms?
A: We think that some of the reports on the death of regulation have been greatly exaggerated.
Regulators have been raising the bar on firms for several years — and their efforts have been focused on detecting the intent to manipulate the market, monitoring manipulation across markets and products, reconstructing trades and uncovering conduct related threats.
These very complex requirements require more sophisticated technology than in the past, which in our case includes the latest in automation and surveillance technologies, leading edge voice and eCommunications analytics, as well as our behavioral approach.
We have also noticed stepped-up attention from compliance officers themselves, since supervisors can now be held criminally liable for certain MAR violations.
Q: How are firms doing as far as compliance with these efforts to fight financial crime?
A: We are finding that financial services organizations are adjusting both the culture of their organization and the technology they use to comply with the new mandates.
Since you need to analyze trade and communications together to detect intent and fully investigate a threat, some firms are combining their communications surveillance and trade surveillance groups into one unified surveillance group.
The front office is taking a more active role in compliance, too.
In some cases, they’re purchasing the same surveillance technology the compliance group uses, but they bring a different eye and knowledge base to investigating alerts. From a technology perspective, big data, machine learning and behavioral analytics are completely changing the nature of surveillance.
Surveillance used to be a narrow box. You used rule-based algorithms to detect known forms of market manipulation by analyzing trade data. Technology is breaking firms out that box. Now you can uncover hidden threats by monitoring the behavior of an individual for anomalies using advanced analytics along with structured and unstructured data.
Q: How viable is holistic surveillance — the marriage of trade and multi-channel communications data — in fighting financial crime?
A: We have firms using holistic surveillance today to detect intent, reconstruct trades and investigate alerts more effectively.
I think you’d be hard pressed to find a CCO [chief compliance officer] that doesn’t believe in the power of holistic surveillance. So clearly, it’s viable.
Maybe the right question is: How widespread is holistic surveillance?
My answer to that question would be that it’s still early days for holistic as firms slowly turn their compliance battleship in a new direction. Accessing internal data — either because you don’t have access to it or you do but it’s not in the right format — remains a challenge.
Some of the organizational challenges I mentioned earlier also remain a challenge. Holistic Surveillance is a journey and you can still see most firms from the dock.
Q: What are some of the latest advances in anti-money laundering (AML) and how are they impacting financial crime?
A: Existing AML programs are suffering from a rising cost of compliance due to high rates of false-positives.
Continuously adding operational headcount has not been the recipe for success, not when the regulatory environment continues to evolve, and the number of products and transactions continue to grow. There is a clear need for a paradigm shift and NICE Actimize has added artificial intelligence [AI] with machine learning and robotics process automation [RPA] to combat this issue.
Machine learning in a transaction monitoring platform will improve true-positives, while significantly driving down false-positives rates. This is applied to optimize customer segments based on like-qualities and risk, not just industry codes, so that highly-targeted models and thresholds can best identify true illicit behavior.
This is used within anomaly detection to detect activity where the typology is not known. Machine learning based predictive scores are also being applied to generated alerts that predict the complexity of the investigation effort and the likelihood that it will result in a suspicious activity report — a technique to help route alerts to investigators in a more efficient manner.
A large part of the compliance cost remains within the human effort to work alerts and cases which can have the proportion of its efforts in the data gathering tasks as opposed to the risk analysis work.
Investigators are being provided visual tools where the upfront workload in triage and investigation is reduced and helps them to uncover the criminal patterns faster.
The workflows are also being infused with AI based on previous outcomes, such as SAR filings, issue generation, and dispositions to suggest flows that will reduce investigation times.
Complementing the machine learning is the integration of robotics to automate manual tasks of investigators. Much of the tedious manual data collection can be done more effectively, more accurately, in a fraction of the time.
Q: Which cutting-edge IT technologies are having the most impact upon financial crime operations and risk management?
A: Artificial intelligence and machine learning, along with behavioral analytics, are transforming compliance and risk management in a number of ways. More specifically, the benefit of entity insights capabilities is one area that stands out for us.
First, this process substantially expands what you can monitor. Instead of being limited to a trade you can now monitor an entity, such as an individual or account, based on profile of that entity.
Second, it enables you to answer the question: Who is putting the firm at risk?
The answer to this question enables firms to proactively investigate threats and mitigate or minimize their damage. Again, holistic is key.
Profiles that use a range of structured and unstructured data will produce a better picture of an entity. Of course, firms still need to be able to answer questions like “did this trade front run another trade” so rule-based algorithms aren’t going to disappear. But being able to identify risk will transform compliance efforts significantly.
Q: How can innovations in electronic communications and voice surveillance help fight financial crime?
A: Natural language processing (NLP)-driven analytics and models are revolutionizing voice and e-communications surveillance.
The previous criticism of e-communications surveillance was that it was inaccurate and generated too many false positives. And for systems that relied on lexicon-based searches that were probably true. However, NLP-driven analytics that understand the context of communications, along with models tuned to recognize the terms and jargon of financial markets, have put those issues in the past. Using this technology, our clients have been able to accurately identify suspicious communications and focus their investigations on true threats instead of false positives.
Q: What can firms do to increase surveillance accuracy and reduce false positives?
A: Trade surveillance models need to be regularly tuned, just like your car.
The models should be tuned to fit your trading activity. After all, a large order for one trader may not be considered large for another. Also, don’t forget about false negatives.
Recently, a large firm was fined $800,000 for setting their alert thresholds so high that the regulator said it would never catch a violation.
For communications surveillance, NLP and understanding the context of communications is essential to reducing false positives.
Finally, correlating trade alerts with communications and behavioral information helps an analyst evaluate an alert better. Since the alert was already generated, this doesn’t technically reduce false positives, but it certainly prevents an analyst from wasting their time chasing one down.
Q: How can behavioral analytics help firms detect conduct risk and intent?
A: Conduct-related threats present a challenge because they don’t follow a predictable path, so you really can’t write an algorithm for them.
Instead, you need to look for anomalies — changes in an individual’s behavior —by analyzing patterns over time.
Behavioral analytics helps you do this.
First you create profiles of “normal behavior” based on an array of risk factors like changes in the number of emails, positions, working hours, number of attachments, and so on.
Then, by using machine learning or statistical models, you can detect anomalies across each risk factor and converting that into a risk score.
Need a Reprint?