FIS, the financial technology company whose systems process transactions for institutions representing nearly 12 percent of global economic activity, announced this week a strategic collaboration with Anthropic to deploy agentic AI directly inside bank operations. The first product of that collaboration is the Financial Crimes AI Agent, built on Anthropic’s Claude model and now in early deployment with BMO and Amalgamated Bank. Broader availability is planned for the second half of 2026.
The agent targets anti-money laundering compliance. A compliance investigation that moves through a bank’s current workflow over the course of several days arrives at the same result in minutes. The system retrieves records, maps relationships between accounts and counterparties, cross-references behavior against known criminal typologies and produces ranked case files for human review. Investigators spend their time on judgment, not assembly. The quality of what regulators receive goes up as the time it takes comes down.
The structure of the partnership is as notable as the product itself. Rather than a conventional vendor relationship, Anthropic’s engineers are working inside FIS, jointly building the system from within the company’s own technical environment. The objective is knowledge transfer: by the time the collaboration matures, FIS will hold the internal capability to design and deploy additional agents without requiring Anthropic’s ongoing involvement. Data governance was non-negotiable in the design. Nothing leaves FIS-controlled systems, and every action the agent takes is logged in a format that can be reviewed and audited.
The program is explicitly designed as a platform, not a point solution. FIS has confirmed that the agent architecture built for compliance will extend into the parts of banking where revenue and relationships are made: credit decisions, new account acquisition, deposit management and fraud prevention. The compliance deployment is the first proof point for an infrastructure play that covers the full customer lifecycle.
The announcement arrived against a backdrop of sustained industry debate about where and how AI agents will reshape banking operations. Yaacov Martin, co-founder and CEO of Jifiti, the white-label lending technology provider, has been among the more direct voices on what that shift will demand of banks. Speaking at FinTech Connect in 2025, Martin described agentic AI as “embedded lending on steroids,” a formulation he has since elaborated in a series of published articles arguing that AI agents are becoming not just tools for banks but the primary channel through which borrowers will discover, compare and execute loan decisions.
Martin’s central argument is that the stakes are not operational but existential. “For the first time in centuries, being a reputable, experienced lender isn’t enough,” he wrote in an article for the Finextra platform. He draws a direct parallel to the early years of internet search. Banks that were not findable were not considered. The same logic, Martin contends, now applies to AI agent ecosystems: if a lending product cannot be discovered and executed by an agent, it does not compete. He has estimated that banks whose systems cannot communicate with AI agents in real time will “lose market share very, very fast.”
The FIS-Anthropic partnership is the most concrete signal yet that the largest infrastructure providers in banking have arrived at the same conclusion. The choice of AML as the first deployment is strategic: compliance work is rules-dense, evidence-intensive and deeply manual, making it a natural fit for agent automation, and a successful deployment there builds institutional confidence before the technology reaches the credit and origination stack. What FIS is building, in effect, is the governance architecture that will make banks ready to trust agents with progressively higher-stakes decisions.
Martin’s $4.28 billion projection for the global AI agents in financial services market by 2032, combined with the $262 billion in AI-influenced consumer sales recorded during the 2025 holiday season, suggests that the demand signal is already present. The FIS-Anthropic announcement is the infrastructure side of that equation beginning to respond.
