Leading financial institutions Goldman Sachs and Deutsche Bank are advancing the use of agentic artificial intelligence to strengthen trade surveillance and compliance efforts. Unlike traditional monitoring systems that depend on preset rules or keyword scanning, these new AI tools dynamically analyze trading patterns as they unfold, flagging unusual activity for human review.
Agentic AI: A New Paradigm in Trade Monitoring
Conventional automated surveillance in banking typically operates on fixed rules — for example, triggering alerts when trades surpass specific sizes or deviate from benchmarks. However, the sheer volume and complexity of modern trading data, spanning multiple asset types and global markets, limits the effectiveness of static approaches. These systems often generate excessive false positives and may miss nuanced manipulative behaviors.
Agentic AI systems aim to overcome these challenges by reasoning through multiple data signals simultaneously and comparing them to historical trading behaviors to detect subtle irregularities. This approach enables the AI to identify complex, previously unrecognized patterns of potential misconduct without relying solely on predefined checklists.
Deutsche Bank’s Collaboration with Google Cloud
Deutsche Bank is actively developing agentic AI solutions in partnership with Google Cloud. This initiative focuses on real-time analysis of extensive order and execution data to promptly surface anomalies. By leveraging generative AI and large language models, the bank extends AI applications beyond customer interactions to sophisticated internal data scrutiny.
The AI agents assess relationships across trades, timing, market dynamics, and trader histories to pinpoint complex anomalies that static systems might overlook. Although these AI tools enhance detection capabilities, final compliance decisions remain the responsibility of human experts.
Goldman Sachs’ Investment in Agentic AI for Compliance
Goldman Sachs has similarly invested in embedding agentic AI within its compliance framework. Building on prior AI deployments in trading and risk management, the bank’s new systems independently scan for behavioral patterns that do not conform to explicit rules but may indicate suspicious conduct.
For regulators, the promise of agentic AI lies in earlier and more accurate identification of market abuse, reducing potential harm and reputational risks. For financial institutions, such AI-driven enhancements help manage the growing volume of alerts efficiently, allowing compliance teams to focus on high-priority cases without sacrificing oversight rigor.
Understanding Agentic AI in Financial Compliance
Agentic AI operates by autonomously selecting relevant data points to analyze, integrating multiple signals, and escalating findings when warranted — all without continuous human prompting. In trading contexts, this means evaluating order flows, price movements, communication metadata, and historical patterns to assess alignment with normal market behavior.
Importantly, these systems do not make disciplinary rulings but serve as advanced tools for information organization and anomaly detection, supporting human decision-making within strict regulatory frameworks.
Implications for the Financial Industry
The rise of agentic AI marks a broader shift towards integrating advanced generative AI architectures in internal controls. Regulatory bodies in the US and Europe urge firms to enhance monitoring of market manipulation, creating incentives to adopt innovative technologies that meet compliance standards more effectively.
However, deploying AI in compliance also introduces challenges such as ensuring model transparency, preventing bias, maintaining data security, and satisfying audit requirements. Banks must carefully govern AI models to uphold regulatory trust and accountability.
If agentic AI tools prove successful, they could redefine compliance workflows by reducing low-value alerts and enabling staff to concentrate on complex, AI-flagged cases. As markets grow faster and more data-intensive, these intelligent systems offer a promising solution to keep pace with evolving risks.
(Image credit: Markus Spiske)
Related Articles
ScaleOps Launches AI Infrastructure Platform to Cut GPU Costs by Up to 70% for Enterprises Running Self-Hosted LLMs
Plaud Introduces NotePin S: A Compact and Affordable Wearable AI Notetaker at CES 2026
U.S. Authorizes Export of Advanced AI Chips to Abu Dhabi’s G42, Strengthening UAE’s AI Ambitions
OpenAI Integrates Codex Capabilities into GPT-5.5, Retires Dedicated Coding Model
