Hawk Outlines Framework for Governing Agentic AI in Financial Crime Investigations
Hawk, the leading provider of AI-supported financial crime and compliance technology, is today publishing a paper setting out a practical governance framework for using agentic AI in anti-financial crime (AFC) investigations.
While regulatory guidance exists for 'traditional' AI - which can predict, classify, and score structured data - agentic AI's advanced ability to plan and execute a sequence of steps to achieve a goal requires extended guardrails, which have yet to be created.
In the new paper, produced jointly with global consulting firm Oliver Wyman, Hawk proposes a framework for extending existing guidance on AI use, thus enabling banks and payment firms to move forward in realizing the significant benefits of using agentic AI to support the investigation of financial crime.
Anti-money laundering (AML) and fraud investigations have traditionally relied on manual effort to review multiple data sources looking for evidence of criminal networks and suspicious activity. Financial institutions are now deploying agentic AI to support that activity, increasing investigation quality while reducing the time needed to complete an investigation.
Governing AFC Agents: How model governance and validation must evolve for agentic AI in financial crime compliance proposes that institutions govern the agentic system as a whole - the model together with its orchestration, tool access, data permissions, and audit trail - rather than the underlying model alone. It introduces a risk-based materiality framework for classifying agents into governance tiers, distinguishes between a use case's inherent materiality and the design choices that shape its residual risk, and sets out validation, monitoring, and vendor-assurance requirements specific to agentic systems.
The paper includes a worked example applying the framework to an AML investigation agent, and a three-stage path that institutions can follow to move from ad hoc oversight to embedded, continuous assurance.
Wolfgang Berner, Chief Product Officer at Hawk, said, "Agentic AI has the power to greatly improve the fight against financial crime. However, agentic AI brings specific governance considerations to anti-financial crime teams, and guidance is not yet available. By extending the work of the Financial Action Task Force and the Wolfsberg Group, we've built a framework to enable financial institutions to move forwards with agentic AI with confidence, in areas such as supporting financial crime investigations."
Stefano Boezio, Partner at Oliver Wyman, said, “Financial institutions have increasingly been exploring the opportunity presented by agentic AI in financial crime compliance. Realizing it safely, however, requires more than technology controls: banks need to evolve governance around how agents are authorized, supervised, monitored, and challenged in day-to-day operations. Those that do so will be positioned to scale agentic AI with confidence."
Governing AFC Agents is available to download here.