Compliance Chiefs Navigate Murky Waters as AI Agents Enter Financial Crime Detection

Compliance Chiefs Navigate Murky Waters as AI Agents Enter Financial Crime Detection

Castellum.AI's CEO outlined the governance challenges facing banks and fintechs as they evaluate agentic AI for anti-money laundering workflows, signaling that regulatory clarity—not just technological capability—remains the primary barrier to adoption.

In a January 21 interview on the Fintech Business Weekly podcast, Peter Piatetsky, cofounder and CEO of Castellum.AI, discussed how compliance leaders should approach the wave of agent-based solutions entering the market, with particular focus on what regulators are signaling about AI adoption and how firms can translate those signals into examiner-ready controls.

The conversation comes as financial institutions face mounting pressure to modernize compliance operations while navigating an uncertain regulatory landscape. Piatetsky's firm builds AI agents specifically for AML and know-your-customer workflows, placing him at the intersection of two forces reshaping corporate finance: the operational imperative to reduce compliance costs and the regulatory requirement to demonstrate control over automated decision-making.

The discussion centered on model governance frameworks for agents—a topic that has vexed compliance officers since large language models entered enterprise workflows. Unlike traditional rules-based systems, AI agents can take actions autonomously, raising questions about accountability, auditability, and the documentation standards regulators will demand during examinations.

Piatetsky addressed how Castellum.AI's agents currently power AML and KYC workflows, though the interview focused more heavily on the governance and regulatory positioning questions than on specific technical capabilities. For CFOs evaluating similar tools, the subtext is clear: the technology may work, but can you explain it to your examiner?

The timing is notable. Financial institutions are simultaneously being told to innovate faster (to compete with nimbler fintech rivals) and to move more carefully (to avoid the compliance breakdowns that have triggered consent orders and executive departures). Agentic AI sits squarely in that tension—promising efficiency gains that could materially impact operating expense ratios, but requiring governance frameworks that most institutions are still building.

Piatetsky also discussed how agents will reshape compliance programs over the coming years, though the interview suggests the transformation will be measured in years, not quarters. The regulatory signaling he described points to a familiar pattern: early adopters will need to invest heavily in documentation and control frameworks, essentially paying a "first-mover tax" to establish what examiner-ready governance looks like.

For finance leaders, the calculus is straightforward but uncomfortable. Compliance costs are real and rising—every basis point matters when investors are scrutinizing efficiency ratios. But so are the costs of getting AI governance wrong, particularly in financial crime compliance where regulatory penalties can dwarf any operational savings.

The podcast episode, hosted by Jason Mikula, is part of a monthly series featuring fintech and banking leaders. The full 52-minute conversation is available on Apple Podcasts, Spotify, and the Substack platform where Fintech Business Weekly publishes.

What remains unresolved—and what Piatetsky's careful framing underscores—is whether the industry will converge on governance standards before or after the first major enforcement action involving an AI agent. That race will determine whether agentic AI becomes a competitive advantage or a cautionary tale in the next wave of compliance transformation.