Feedzai launches Farol to connect banks’ agentic AI

Feedzai has introduced Farol, software that sits between agentic AI models and banks’ operational systems to connect conversational and decision-making agents to existing fraud, risk and compliance platforms. The product is designed to let institutions use agents for workflows without replacing back-end systems or exposing sensitive data.

Farol provides prebuilt connectors to payments engines, transaction monitoring, case management and identity services. The layer applies controls that limit what agents can query and which actions they can take, and it records agent actions and decisions for later review.

The product includes role-based access controls, detailed logging and query filtering to limit customer-data exposure. Developers can map business rules and compliance policies into agent behavior using low-code tools. Feedzai says the platform captures step-by-step decision records and explainability information to support audits and regulatory reporting.

Farol supports human-in-the-loop workflows so analysts can accept, modify or reject agent recommendations. Monitoring tools track agent performance, error rates and drift in decision outcomes, which teams can use to adjust models and policies over time.

Deployment options include cloud and hybrid configurations to accommodate institutions that keep sensitive data on-premises. The vendor positions Farol as compatible with multiple large language models and agent frameworks and adds a policy layer to mediate model access to data.

Feedzai outlines practical use cases such as automating investigation workflows by summarizing transaction histories, handling customer requests that require risk checks like transaction disputes, and assisting analysts by retrieving relevant rules and precedents during complex cases. Documentation states Farol is meant to integrate with existing workflows and compliance controls.

Founded in 2011, Feedzai has a history in fraud detection and risk management for banks and payment processors. The company said customers that already use its detection products may be able to adopt Farol with fewer integration steps.

Regulators are increasing scrutiny of AI in finance, and banks deploying agents must document controls, data flows and decision logic. Feedzai said Farol’s logging and governance features are intended to support those documentation needs. Financial institutions assessing agent platforms will compare governance features alongside model performance, cost and integration effort.