Feedzai has launched Farol, an AI agent embedded in its RiskOps Studio that supports fraud investigations and rule analysis for banks.
The agent answers questions in plain language, analyses rule performance automatically, and produces case summaries within the software analysts already use.
The launch comes as the cost of fighting fraud rises on several fronts. Feedzai states that the mandatory reimbursement rules are increasing banks' exposure to losses. At the same time, spending on technology and manual review keeps growing. Farol is designed as a way for fraud teams to manage more work without adding staff.
The company also points to a gap between testing agentic AI and getting results from it. Feedzai’s research indicates that 68% of financial institutions are actively testing agentic AI. However, most deployments do not deliver operational efficiencies because third-party AI models operate separately from real-time transaction data. Farol is intended to close that gap by working directly inside existing risk workflows. There, it has full context, keeps transparent audit trails, and can carry out tasks on its own without changes to core infrastructure.
Four initial skills
Farol launches with four purpose-built skills. The first, Risk Strategy, examines rule sets to find patterns that generate alerts without catching fraud and recommends revised thresholds. According to Feedzai, this cuts rule hygiene work from days to minutes. The Investigations skill retrieves and summarises alert data, so analysts have the context they need to work through cases more quickly. The company reports that it has reduced alert handling times by 20%.
The other two skills focus on support and reporting. Knowledge answers users' questions about how to conduct tasks on the platform without them leaving the interface. SAR Drafting compiles and summarises the information needed for Suspicious Activity Reports (SARs), which Feedzai says allows these reports to be drafted up to 12 times faster.
The agent runs inside each financial institution's own environment. This means the information and insights it produces stay within the customer's estate rather than moving to an external system.
Farol builds on Feedzai's earlier work in financial crime prevention. That work includes putting machine learning into production, explainable model outcomes, real-time risk decisions across channels, and federated network intelligence.
Industry perspectives
Sam Abadir, Research Director, Risk, Financial Crime, and Compliance at IDC, described fraud investigation as one of the clearest near-term use cases for agentic AI. In the analyst's view, an agent with direct access to case data and analyst tools removes the manual work of assembling information, so cases are resolved faster without extra headcount.
For fraud operations, the release describes a shift in how AI is being deployed. Instead of standalone models, agents are being placed inside the platforms where transaction data, rules, and case work already sit.
