Financial institutions are advancing their use of artificial intelligence to ensure market integrity, as leading banks such as Goldman Sachs and Deutsche Bank test adaptive “agentic” AI for trading surveillance. These dynamic systems are being trialed to detect unusual trading patterns across vast volumes of data, aiming to strengthen regulatory compliance. As financial markets become more complex, these tools offer potential solutions to the long-standing challenge of efficiently identifying misconduct while managing a growing flood of transactional information.
Goldman Sachs and Deutsche Bank have each explored algorithmic surveillance in the past, primarily with technologies that flagged orders based on predetermined triggers. The latest shift involves systems that reason through data in real time, departing from earlier, rule-based approaches. While traditional methods helped firms comply with regulatory expectations, they often produced excessive false positives and could miss subtle forms of manipulation. The current initiative builds on these experiences and seeks to combine higher accuracy with adaptive monitoring capabilities.
How Does “Agentic” AI Change Surveillance?
Unlike previous automated surveillance tools that relied strictly on static rules, agentic AI evaluates multiple signals and historical contexts simultaneously. These agents can assess clusters of trades, scrutinize conduct patterns, and raise alerts when they encounter behavior that does not match established norms. The outcome is a narrowed focus on genuinely unusual cases, enabling compliance teams to direct their efforts where oversight is most needed.
What Are Deutsche Bank and Goldman Sachs Doing?
Deutsche Bank has collaborated with Google Cloud to introduce AI agents that monitor order and trade data for signs of irregular activity. The bank’s compliance staff are still responsible for investigating anomalous cases surfaced by the system. Speaking about these advancements, a Deutsche Bank representative stated,
“We’re using cutting-edge AI to process large sets of trading data and enhance our ability to detect suspicious activity.”
Meanwhile, Goldman Sachs has integrated agentic AI into its surveillance efforts, targeting patterns that evade traditional detection.
Can Artificial Intelligence Replace Human Oversight?
AI tools, while capable of filtering and ranking alerts, do not replace the decision-making roles of compliance professionals. Instead, these systems act as a supplement by organizing information for more efficient human review. A spokesperson from Goldman Sachs commented,
“AI agents help us identify conduct risks more efficiently, but human judgment remains essential.”
Regulators are keeping a close watch to ensure that these new technologies maintain transparency and fairness in compliance processes.
Market watchdogs in Europe and the US do not yet require agentic AI, but they have pushed firms to improve surveillance capabilities. As banks consider broader adoption, concerns persist around transparency, data security, and the interpretability of results. Previous announcements about AI in finance showcased its use in front-office decision-making and customer service, but the current emphasis is on compliance and internal controls. Unlike earlier experiments that focused on client interaction, this wave centers on analyzing the microstructure of markets.
Banks investing in agentic AI surveillance are betting such systems will help reduce operational noise and focus attention where it counts. These developments underscore a shift toward harnessing technology not only for efficiency, but also for regulatory robustness. Stakeholders throughout the financial ecosystem—from regulators to internal auditors—will be monitoring the progress of these trials closely. Those considering similar systems should prioritize model governance, auditability, and the ongoing involvement of experienced compliance staff to preserve trust in automated oversight.
