FIS has made its AML Investigations Agent commercially available to all US-based banks, including those that are not existing clients.
Developed with support from Anthropic and design partner Amalgamated Bank, the AI-powered agent aims to reduce the manual work involved in financial crime investigations. Amalgamated Bank is now deploying the tool.
The launch follows FIS’s initial announcement with Anthropic in May 2026.
FIS projects substantial reductions in AML investigation time
FIS cited 2023 LexisNexis estimates showing that US financial institutions spend US$35 billion to US$40 billion annually on AML operations.
According to 2025 research from AnChain.AI, analysts spend 40% to 60% of investigation time gathering information before analysis can begin. Datos Insights estimated industry false-positive rates of 90% to 95% in 2023.
Based on Amalgamated Bank’s existing workflows and the agent’s expected capabilities, FIS projects a reduction of more than 90% in Level 1 alert review time and approximately 98% in Level 2 case investigation time.
Time spent preparing suspicious activity report (SAR) narratives could also fall by more than 80%.
These figures are projections rather than measured results from the bank’s deployment.
How the AI agent handles investigations
The agent evaluates alerts against each customer’s profile, historical transactions and expected behaviour before gathering supporting evidence.
FIS says this approach helps distinguish routine activity from potentially suspicious transactions, such as cash deposits deliberately kept below federal reporting thresholds.
Each finding links to its source record, document or cheque image. The system also records recommendations, supporting reasoning and final analyst decisions in an audit trail for internal reviews and regulatory examinations.
FIS built the agent on its Trust Assured framework, which incorporates governance, compliance and responsible AI controls.
Banks retain their own risk thresholds and investigation procedures through the FIS Control Plane, which manages agent access and institution-specific configurations.
The agent works alongside existing monitoring and case management systems rather than replacing them. It integrates with FIS Horizon and FIS Systematics, with FIS IBS support planned.
“Investigations don’t slow down because analysts lack judgment. They slow down because evidence is scattered across systems and every alert gets treated the same,”
said Stan Szwalbenest, Head of Enterprise Fraud and Financial Crimes at FIS.
Built using Anthropic’s Claude models
FIS developed and owns the agent, which runs in its private cloud using Anthropic’s Claude models.
An Anthropic team worked alongside FIS product and compliance teams during development, helping establish evaluation frameworks and engineering practices for future AI agents.
Commercial availability currently covers US-based banks.
Featured image credit: Edited by Fintech News Singapore, based on image by mrdon via Magnific
