By FISLearning Objectives
- Understand how an AI agent can compress anti-money-laundering investigations from days to minutes
- Learn what agentic financial-crime investigation looks like end to end, from evidence gathering to human decision
- Recognize why false positives, explainability, and human accountability remain central to AML work
What Is FIS Financial Crimes AI?
FIS Financial Crimes AI is an agentic anti-money-laundering (AML) tool that automates the slow, manual parts of a financial-crime investigation. Instead of an analyst spending hours pulling data from disconnected systems, the agent assembles evidence across a bank's core systems, evaluates the activity against known money-laundering typologies, drafts a case disposition, and surfaces the highest-risk alerts for an investigator to decide. The goal is to compress investigations that once took days into minutes, cut false positives, and improve the quality of suspicious-activity-report narratives — the formal filings banks submit to regulators.
The product is built in partnership with Anthropic on its Claude models. Anthropic's applied-AI team and forward-deployed engineers are embedded with FIS to co-design the agent, pairing Claude's reasoning with FIS's banking data and regulatory infrastructure. Crucially, it runs in an agent-first governed environment where client data stays within FIS-controlled infrastructure and every agent decision is traceable and auditable.
Fidelity National Information Services (NYSE: FIS), founded in 1968 and headquartered in Jacksonville, Florida, is one of the largest financial-technology providers to banks worldwide. It announced the Financial Crimes AI Agent in May 2026 as the first step in a broader plan to bring agentic AI across banking workflows, with early deployments at institutions including BMO and Amalgamated Bank and wider availability planned through the second half of 2026.
💡Key Concept
Agentic AML investigation: In a traditional alert review, an investigator manually gathers transaction records, customer details, and prior cases from several systems before any analysis begins. An agentic approach flips the order — the AI agent collects the evidence, compares it against known laundering patterns, and drafts a recommended disposition, then a human investigator reviews the reasoning and makes the final call. The agent does the assembling; the person keeps the decision.
✅Tip
Visit FIS Financial Crimes AI: fisglobal.com — for banks and financial institutions modernizing AML operations; enterprise platform, pricing by arrangement.
Core Capabilities
Automated evidence assembly
The agent gathers the transaction history, customer information, and related cases needed to investigate an alert, pulling from a bank's core systems so the investigator does not have to stitch the picture together by hand.
Typology evaluation
It compares the assembled activity against known money-laundering typologies — recognized patterns of illicit behavior — to judge how closely a case resembles genuine financial crime rather than benign activity.
Drafted disposition and narratives
The agent proposes a case disposition and drafts suspicious-activity-report narratives, giving investigators a strong first draft that improves consistency and reduces the time spent writing up filings.
Governed, auditable environment
FIS runs the agent in a governed setup where client data stays within its controlled infrastructure and every decision is traceable, supporting the audit trail regulators expect.
Strengths
- Speed on a real bottleneck: Automating evidence gathering compresses investigations from days to minutes, the step where analysts lose the most time.
- Fewer false positives: Better pattern evaluation helps focus investigator attention on cases that genuinely warrant review.
- Higher-quality filings: Drafted disposition and narrative support improves the consistency of suspicious-activity reports.
- Built for banking constraints: Data stays in FIS-controlled infrastructure with traceable decisions, addressing the regulatory bar from the start.
Limitations & Considerations
- Humans keep accountability. The agent drafts and recommends, but investigators and the institution remain responsible to regulators for every filing and decision.
- False positives and explainability are the core challenge. AML systems must justify why an alert matters; an agent's reasoning has to be clear enough for investigators and examiners to trust and defend.
- Accuracy and hallucination risk. As with any language-model system, outputs can be wrong or overstated, so human review of the evidence and disposition is essential, not optional.
- Early, staged rollout. The agent launched with a small set of institutions and is expanding through 2026, so broad, long-run performance is still being established.
Best Use Cases
| Task | Why FIS Financial Crimes AI |
|---|---|
| Investigating AML alerts | Assembles evidence and drafts a disposition in minutes |
| Reducing investigator backlog | Surfaces the highest-risk cases first |
| Writing suspicious-activity reports | Drafts narratives to improve quality and consistency |
| Meeting audit requirements | Runs in a governed, traceable environment |
Getting Started
- Assess your bank's current AML alert volume and where investigators spend the most manual effort.
- Engage FIS to understand deployment within its governed environment and how data stays in FIS-controlled infrastructure.
- Pilot the agent on a defined alert queue, keeping investigators in the loop to review every drafted disposition.
- Measure investigation time, false-positive rates, and filing quality before expanding to more workflows.
Key Takeaways
- FIS Financial Crimes AI is an agentic anti-money-laundering tool built with Anthropic on Claude, launched in May 2026.
- It gathers evidence across core systems, evaluates typologies, and drafts dispositions and suspicious-activity narratives, compressing investigations from days to minutes.
- The agent assembles the case, but a human investigator keeps the decision, and every step is traceable and auditable.
- The honest caveats are that false positives and explainability remain the hard part of AML, and humans stay accountable to regulators.
