The financial sector is undergoing a significant transformation with the advent of agentic AI. These autonomous systems are no longer confined to predictive tasks; they now reason, decide, and act independently. This shift is redefining governance and compliance frameworks, as traditional models struggle to keep pace with the dynamic nature of agentic AI.
From underwriting assistants to fraud detection workflows, these systems are becoming integral to financial operations. However, their autonomous nature poses unique challenges for governance and compliance, requiring a fundamental rethinking of existing frameworks.
The Evolution of AI in Financial Services
Agentic AI represents a paradigm shift from traditional AI systems. Unlike static models that provide single outputs for given inputs, agentic AI can break down tasks into multiple steps and execute them autonomously. This capability is revolutionizing various aspects of financial services, including underwriting, anti-money laundering (AML) investigations, and fraud detection.
For instance, underwriting assistants can now gather data, weigh factors, and recommend next steps autonomously. AML investigation agents can triage alerts and assemble case files, while fraud detection workflows adapt to new patterns in real time. These advancements are enhancing efficiency and accuracy but also introduce new governance and compliance challenges.
Governance and Compliance Challenges
The primary challenge lies in the dynamic and autonomous nature of agentic AI. Traditional governance frameworks were designed for static models with predictable outputs. In contrast, agentic AI systems behave dynamically, making it difficult to ensure consistent behavior and accountability.
One of the key issues is the lack of visibility into autonomous decisions. Teams often struggle to understand why a decision was made, trace the full path of actions taken by an agent, and monitor how its behavior shifts over time. This lack of transparency complicates governance and auditing processes.
Regulatory expectations are also evolving. Supervisors increasingly demand explainability, clear accountability, ongoing monitoring, and durable governance evidence. Established guidelines, such as the SR 26-2 in the US and OSFI E-23 in Canada, emphasize continuous, function-based risk management, which aligns with the principles needed for governing agentic AI.
Compliance Gaps and Risks
The transition to agentic AI exposes several compliance gaps. Traditional approaches to validation, monitoring, documentation, and accountability are no longer sufficient. Continuous review, behavioral oversight, real-time traceability, and multi-team governance are essential to address these gaps.
Regulatory compliance risks are particularly concerning. Undocumented decisions, actions that breach policy, and the inability to explain decisions after the fact can lead to significant regulatory exposure. Financial institutions must adapt their governance models to ensure that autonomous systems remain effective and accountable.
First National Bank of Omaha’s Experience
First National Bank of Omaha (FNBO) has emerged as an early adopter of agentic AI for financial crime investigations. The bank reports a 50% reduction in investigation time, thanks to AI agents handling repetitive tasks and providing consistent responses to fraud, money-laundering, and sanctions-violation alerts.
Nick Baxter, Chief Risk Officer at FNBO, highlights the consistent challenges of financial crimes and the need for better solutions. Chuck Subrt, Director of the Fraud and AML Practice at Datos Insights, notes that agentic AI is gradually moving beyond pilot projects into full-scale financial crime operations. These agents are completing high-volume, repetitive tasks, allowing human investigators to focus on more complex analyses.
The deployment of agentic AI in financial crime investigations is part of a broader trend. As financial crime continues to rise, with illicit activities amounting to trillions of dollars annually, the need for efficient and effective investigative tools becomes paramount. AI agents are proving to be valuable assets in this fight.
