Banks Are Building AI Watchdogs 🤖🤯
AI
🎧
Banks are currently evaluating a new type of artificial intelligence, often described as “agentic AI,” for use in trading surveillance. Systems are being tested that move beyond simply scanning for keywords or adhering to pre-set rules. Trading desks are utilizing software agents designed to analyze patterns in real-time, flagging activity for human review. These agentic surveillance tools monitor trading activity, triggering alerts if a trade surpasses a defined size, deviates from a benchmark, or matches a recognized risk pattern. Compliance teams then manually examine these alerts. Regulators in the US and Europe have encouraged firms to bolster the monitoring of market abuse and manipulation. The deployment of this technology represents an effort to enhance the scrutiny of trading activity and identify potential misconduct.
AGENTIC AI: A NEW FRONTIER IN FINANCIAL SURVEILLANCE
Banks are experimenting with a novel approach to trading surveillance – agentic AI – that transcends traditional keyword scanning and pre-defined rule sets. Instead of relying solely on static alerts, some trading desks are utilizing systems designed to reason through patterns in real-time and flag conduct requiring human review. This shift represents a significant evolution in how financial institutions monitor trading activity and mitigate risk.
THE CHALLENGE OF SCALE AND COMPLEXITY IN TRADING SURVEILLANCE
Traditional trading surveillance systems often rely on automated alerts triggered by predefined rules – such as trades exceeding a certain size, deviating from a benchmark, or fitting a known risk pattern. Compliance teams then manually review these alerts. However, modern markets generate massive volumes of data across diverse asset classes, time zones, and trading venues. This complexity leads to a high volume of false positives, while subtle forms of manipulation may not align with established patterns. The static nature of these systems struggles to cope with this scale, presenting a significant operational hurdle.
AGENTIC AI: A SYSTEMATIC APPROACH TO ANOMALY DETECTION
The newer agentic systems aim to move beyond this traditional approach. Rather than simply matching trades against a checklist, AI agents are designed to examine trading behavior across multiple signals, compare it with historical activity, and detect unusual combinations of actions. These tools are not intended to replace compliance officers but rather to function as an additional layer of monitoring, surfacing cases that warrant closer human inspection. The focus is on identifying “complex anomalies” in orders and trades, considering relationships between trades, timing, market conditions, and trader history, rather than isolated events.
GOOGLE CLOUD AND DEUTSCHE BANK: PIONEERING AGENTIC AI
Deutsche Bank is actively collaborating with Google Cloud to develop AI agents specifically designed for trading activity monitoring. The system analyzes large sets of order and execution data, flagging anomalies in near real-time. This initiative reflects Deutsche Bank’s broader commitment to applying generative and large language model technology – beyond customer-facing interfaces – to internal control functions. Goldman Sachs is also exploring agentic AI for surveillance, building on its substantial investments in AI across its trading and risk systems.
UNDERSTANDING THE CORE PRINCIPLES OF “AGENTIC AI”
The term “agentic AI” refers to systems capable of taking goal-directed actions, rather than simply responding to prompts. This means the software can autonomously decide what data to examine next, compare multiple signals, and escalate findings without constant human intervention. In a trading context, this might involve monitoring order flows, price movements, communications metadata, and historical behavior to assess whether activity aligns with normal patterns. Crucially, this does not imply the system makes disciplinary decisions independently. Financial institutions operate under strict regulatory regimes, and accountability remains with human supervisors. The agent’s role is to identify and organize information more effectively than static systems can.
AGENTIC AI AS PART OF A WIDER COMPLIANCE SHIFT
The application of more advanced generative AI architectures to internal control functions represents a broader shift in the compliance landscape. Regulators in the US and Europe have encouraged firms to improve the monitoring of market abuse and manipulation. While rules do not mandate agentic AI, they do require firms to maintain effective systems and controls. If AI tools can help meet this standard, adoption is likely to grow.
KEY CONSIDERATIONS FOR IMPLEMENTATION
The successful deployment of agentic surveillance tools raises several critical questions. Banks must ensure that the underlying models are explainable, that they do not introduce bias, and that they can withstand rigorous regulatory review. Model governance, data security, and comprehensive audit trails remain central concerns. The ability to analyze patterns in real-time is becoming increasingly difficult to achieve with rule-based systems alone, particularly in markets characterized by rapid data volume growth.
POTENTIAL IMPACT ON COMPLIANCE WORKFLOWS
If agentic surveillance tools prove effective, they could fundamentally alter how compliance teams operate. Instead of sifting through large volumes of simple alerts, staff may dedicate more time to evaluating complex cases surfaced by AI agents. This shift would not eliminate the need for human judgment, but it could redirect human effort toward more strategic analysis. The future of compliance may involve a collaborative approach, where AI agents highlight potential issues, and human experts provide the critical assessment and decision-making.
This article is AI-synthesized from public sources and may not reflect original reporting.