Trade Surveillance and AI
Trade Surveillance and AI
A Tale of Two Industries
Over the past month, I've engaged in fascinating discussions with surveillance leaders across Asia-Pacific, meeting with teams in Singapore and Hong Kong in May, followed by conversations with the commodities surveillance leaders in Europe and the UK in early June. While these groups face different pressures, commodities teams grappling with REMIT II implementation and ACER's enhanced enforcement powers, while APAC teams focus on data remediation and growing regulatory divergence and the tariff-induced volumes, an intriguing theme emerged that I found both surprising and distinctly different from bulge-bracket bank surveillance leaders.
The Communications Success Story
There's resounding acceptance across the industry that AI's impact on communications surveillance has been immense in just a few short years. Teams are deploying cutting-edge technology with embedded AI in various forms, seeing benefits ranging from effectiveness in scope (particularly languages) to an increasing range of risk detection capabilities and efficiency gains through false positive reductions and expanded population coverage.
While nobody is declaring "gold standard" status yet, a wide variety of institutions are experiencing daily benefits in surveillance scope, quality, and cost. The transformation has been remarkable across institutions of all sizes and scales.
Trade Surveillance: The Missing Revolution
However, trade surveillance tells a markedly different story across two distinct groups of surveillance leadership.
Large banks and asset management firms in European and US financial centres are familiar with, and increasingly deploying, AI and machine learning in their detection strategies across multiple asset classes. We've seen various vendor technologies implementing these capabilities, while some better-resourced technology teams are building aspects internally.
In stark contrast, both banking and asset management regional surveillance leaders in APAC, as well as the commodities sector leaders in Europe, are not following the same path. The benefits and impact of AI haven't reached the market at scale and the commitment and enthusiasm achieved in communications surveillance is totally missing here.
The Paradox
This disparity is particularly baffling given the circumstances. Trade surveillance teams operate under far greater regulatory requirements and high-profile enforcement examples, while facing equal internal pressure to manage costs and improve coverage. With the potential for AI to enhance both efficiency and effectiveness, trade surveillance should be a natural winner.
Consider the fundamentals: highly structured datasets, phenomenal market volatility over the past five years, and rule-based detection systems with thresholds and parameterization creating a never-ending cycle of adjustments, incredibly high false positive volumes. The opportunity for significant improvements is compelling.
The Reality Check
Yet in recent discussions, I encountered a different reality. While some larger firms talk confidently about AI's impact on trade surveillance; acknowledging it's more complex and potentially expensive but showing real commitment to leveraging AI; I was met with responses that ranged from resignation to scepticism:
"We aren't ready for AI yet"
"It isn't the same as communications"
"We don't know how to do this and we're wary of model risk management, so we'll wait and see"
"There really isn't regulatory pressure to move to AI solutions in the trade space"
Sections of this group really couldn’t see what AI might be able to do, for the efficiency or effectiveness of their programs.
Understanding the Divide
Several factors may explain this divergence.
Program maturity varies widely between these groups. Commodities teams trading physical commodities face immediate challenges implementing new REMIT II standards this year, requiring focused onboarding of numerous venues and additional datasets, with calibration of the risks presented in specific auctions and the impact of 15-minute intervals. Perhaps AI solutions seemed too complex when "day-one" compliance was the necessity.
Many institutions globally are still grappling with cross-product coverage, and few vendors truly understand or offer comprehensive solutions to this risk, leaving firms to solve these challenges internally.
APAC teams, often regional banks with lead regulators taking longer to push the agenda forward, may not face the same pressure as their Western counterparts, though the demand for high-quality, efficient surveillance should be similar.
The Budget Reality
One significant difference lies in budgets. While AI in electronic communications has been positioned as a cost-reduction strategy, in trade surveillance it's perceived as an expensive solution. This perception gap may be crucial in adoption decisions.
Vendor Community Division
Another key difference is the vendor community itself. Trade surveillance providers remain divided on AI's role, with some vendors adhering strictly to traditional rule-based strategies without any AI or machine learning deployment, while others have fully embraced AI as the path forward.
This contrasts sharply with communications providers, who are largely incorporating these tools into their platforms in some form or another, and messaging on this. Few in trade surveillance are promoting a blended approach, and banks aren't easily adopting the hybrid strategies they've successfully implemented in communications.
The trade surveillance vendor space appears to be less driven by innovation at the moment and more what resembles a belief system; firms are either in the "pro-AI camp" or they're not. Does this divide potentially also represent the surveillance practitioners themselves?
However, a regulatory catalyst for changing all this may have appeared.
Into this dynamic, last week at London Tech Week, stepped the FCA, announcing its collaboration with NVIDIA to provide a supercharged sandbox for firms to safely test AI, underlining the regulator's commitment to the technology.
Perhaps this regulatory endorsement will shift trade surveillance ecosystem toward a commitment to finding the most effective and efficient ways to use AI to detect market misconduct.
Looking Ahead
The contrast between communications and trade surveillance adoption of AI raises important questions about innovation diffusion in financial services. While one area has embraced transformation, another remains cautiously divided despite facing similar pressures and potentially greater opportunities.
As surveillance practitioners continue to navigate these technological decisions, the question isn't whether AI will eventually transform trade surveillance, but how quickly the industry can move toward practical implementation.
As someone who regularly works with surveillance teams navigating these complex technology decisions, I find this divergence fascinating and concerning. The potential benefits are clear, the regulatory support is growing, and the operational pressures remain intense. If you're evaluating your trade surveillance strategy and considering how emerging technologies might enhance your capabilities, the landscape is evolving rapidly, and the right approach today may look very different from conventional wisdom.
