"Every bank in the world wants AI that acts, not just assists," FIS CEO Stephanie Ferris said, framing a sales pitch that pairs a major financial-technology vendor with a leading AI model builder to tackle one of banking's most expensive compliance headaches.
The Financial Crimes AI Agent
Fidelity National Information Services (FIS) and Anthropic have built a product called the Financial Crimes AI Agent that uses Anthropic's Claude model as its reasoning engine. According to FIS, when a case opens the agent automatically pulls evidence across a bank's core systems, compares activity against known patterns of illicit behavior and presents investigators with a ranked case file. FIS says the model can do in minutes what currently requires days and plans to offer the capability for general availability planned in the second half of this year.
FIS and Anthropic: embedded engineers and a governance pitch
The partnership places Anthropic engineers embedded inside FIS to build the agent, a setup Anthropic's head of financial services, Jonathan Pelosi, said is designed to ensure traceability. "Every conclusion the agent reaches links back to its source data, and every decision stays with the investigator," Pelosi said. FIS is marketing itself not as a standalone tool vendor but as the governing layer — "a trusted provider who manages the data, who governs the agents, and who stands between your customers and the AI making decisions about their money," Ferris said — positioning the company as an intermediary that banks would use to access agent capabilities rather than integrating directly with an API provider.
Traceability, auditability and regulatory accountability
The pitch emphasizes traceability and knowledge transfer. FIS says the knowledge transfer component is structured so FIS can eventually build and scale additional agents without Anthropic's ongoing involvement, giving FIS a degree of independence that pure API integrations typically don't. But regulatory consultants at Dutch multinational Wolters Kluwer warned that "regulators have made clear through guidance, examinations, and enforcement actions that automation does not reduce accountability." The source material also observes that autonomous AI agents in financial crime workflows must operate within auditable frameworks, providing a clear reasoning chain for every decision in order to satisfy transparency and human oversight requirements, and it remains unclear if FIS' architecture satisfies the bar for regulators in practice and not just in design documents.
Scale of the problem: dollars, false positives and illicit flows
The announcement arrives against a backdrop of large numbers. FIS describes itself as powering nearly 12% of the global economy. Financial crime compliance costs in the United States and Canada surpass $61 billion a year, according to the report. The United Nations estimates that up to $2 trillion in illicit funds moves through the global financial system annually. Robert Edwards, former global head of transaction monitoring at Credit Suisse, argued that legacy rules-based engines have produced "99% false positives" and that AI "gives us an opportunity to become much more effective in our transaction monitoring and alert resolutions."
What this means for technologists, regulators, and banks
- Technologists and security teams: The agent model centers traceability — every conclusion linked back to source data — and a knowledge-transfer plan so FIS can scale agents internally. Engineers will be watching whether those technical assurances translate into auditable logs and demonstrable human oversight that examiners accept.
- Regulators and compliance examiners: Wolters Kluwer's consultants reiterate that automation does not reduce accountability. Examiners will expect banks deploying agents to demonstrate that personnel understand how the model reaches conclusions and that decisions remain auditable and overseen by humans.
- Banks and procurement leaders: FIS positions itself as the governing layer between customers and AI-driven decisions and plans to expand the platform beyond financial-crime investigations to credit decisions, deposit retention, customer onboarding and fraud prevention — all built on the same platform and powered by Claude. Banks will need to weigh potential speed gains (minutes vs. days) against regulatory, operational and governance requirements.
FIS and Anthropic present a concrete answer to a familiar problem: large compliance costs, many false positives and slow investigator workflows. The partnership promises speed, traceability and an on‑premise governance posture that FIS argues will be attractive to banks. But the facts in the announcement also make clear what remains to be seen — whether the design assurances and embedded-engineer model will meet the practical test of regulatory examiners and whether adoption will live up to the projected reduction in time and false positives.
