TECH TUESDAY: Integrating Gen AI in Surveillance

TECH TUESDAY is a weekly content series that covers all aspects of capital markets technology. TECH TUESDAY is produced in collaboration with Nasdaq.

Nasdaq recently announced the integration of generative AI in its Market Surveillance technology, which is used by regulators and marketplaces worldwide. Traders Magazine sat down with Tony Sio, Head of Regulatory Strategy and Innovation, to learn more about the capability and its applications.

Is this the first or most significant utilization of gen AI in market surveillance?

Yes, we believe that this is the first time that generative AI has been properly incorporated into a market surveillance tool. However, this isn’t a rushed job. We’ve been working on this since 2023. We methodically worked it from ideation to proof of concept (PoC) to product, each step of the way ensuring that what we are doing directly solves the problems faced by surveillance analysts.

Can you describe the problem you’re solving and how generative AI helps?

Our clients’ market surveillance teams routinely get hundreds of alerts daily; some are false positives and others require a deeper dive. For each anomaly, staff then need to understand the context in which it was triggered, and generally, they are for smaller, lesser-known stocks. For example, there may be takeover rumors or product launch details buried in a corporate filing; at the same time, most stocks that analysts look at aren’t well-known and they may have no background of that stock. It’s time-consuming for analysts to find what they need and even worse if they miss something critical. The goal of what we created is to streamline that process using the best of generative AI combined with our existing surveillance tools.

How does this enhanced surveillance product compare with other market offerings? Do you believe it’s the best product on the market?

Nasdaq Market Surveillance has been and still is the best market surveillance offering. We’ve stayed in the top spot by doing two things. Firstly, we are razor-focused on solving the problem of keeping markets safe; secondly, we are very proactive in embracing new technologies. What this means is that we are constantly experimenting and bringing new capabilities. However, we aren’t experimenting just for the sake of it; everything needs to be for the purpose of keeping markets safe. This discipline and focus mean that we aren’t just chasing the shiny new thing. If it isn’t helping to solve the core problem, we quickly move on. This philosophy is why many regulators and exchanges are serious about tackling market abuse and partnering with us on technology.

What are clients saying about this enhanced surveillance offering – is it something they’ve been asking for?

Clients have been extremely positive. We are trying to relieve them of the mundane manual tasks so that they can focus on more creative and deeper analytics. During the initial PoC, we received feedback indicating an up to 30% reduction in time for some of the investigative steps analysts needed to take. Even some of the simpler parts of the tool reduced the number of clicks and number of documents they would need to sift through, which translates into immediate time savings. Time that could be spent on much more creative tasks.

What considerations did you need to address for this specific problem?

We work in one of the most highly regulated industries in the world. When we started the project, we sat down with the analyst community to draw up the guardrails. The No. 1 thing flagged was that we needed to ensure the accuracy of the information we provided. On top of that, we wanted all text produced by the tool to be traceable to its source. This has limited what we could do to some extent. Given the regulatory environment and nature of the problem, however, it was understandable, and we built along those lines. We were highly focused on avoiding any misleading or fake news, not to mention hallucinations or anything of that type, in the product itself. The key is designing for your specific use case. We are considering incorporating additional sources, such as social media, for future iterations without diluting the initial guardrails. We believe this is possible through thoughtful design and user transparency.

How else might gen AI be used in surveillance?

We see this as just the first step to incorporating gen AI even deeper into the product; the capabilities of the technology continue to improve even as we integrate it. I see two immediate areas where we could use it further. Firstly, gen AI could be used to better incorporate unstructured signals into the market monitoring process, such as sentiment, by combining a broader set of signals into the monitoring program. Therefore, we will be able to encapsulate proper market behavior better and detect misbehavior. Secondly, I feel that we have only started on the journey of improving the investigation workflow with gen AI. As part of our deployment of the tool, we are working very closely with our customers and Nasdaq’s own surveillance team to improve the tool and address a broader range of activities. I even hope that the time freed up by using these new tools is then used by the analysts in new creative investigation techniques, which we can improve even further with gen AI.

Originally published on Traders Magazine

For more on the business of trading, subscribe to Traders Magazine’s weekly newsletter here.