CASE: An Agentic AI Framework for Enhancing Scam Intelligence in Digital Payments

Google researchers have developed CASE, an AI framework using conversational agents to collect detailed scam intelligence from potential victims across digital payment platforms. Implemented on Google Pay India, the system increased scam enforcement actions by 21% by extracting structured data from victim interviews to identify sophisticated social engineering patterns.

Digital payment platforms face an escalating challenge: while their growth has democratized commerce globally, it has simultaneously created a lucrative target for sophisticated scammers who orchestrate fraud across multiple channels outside the platform itself. Traditional transaction monitoring and user-behavior signals prove insufficient because they lack context about how scams are initiated, coordinated, and executed. CASE addresses this intelligence gap by deploying conversational AI agents that proactively interview potential victims, transforming unstructured victim accounts into actionable structured data.

The framework represents a methodological shift in fintech security. Rather than purely algorithmic detection, CASE combines human-centered interviewing through AI with downstream machine processing, recognizing that victims often possess critical intelligence about scam methods that automated systems cannot capture. This hybrid approach acknowledges the sophistication of modern social engineering, which frequently combines multiple communication channels, psychological manipulation, and coordinated timing to bypass technical controls.

The 21% enforcement uplift observed on Google Pay India demonstrates measurable real-world impact, suggesting that deeper scam intelligence directly translates to faster, more accurate intervention. This outcome carries implications for the broader fintech ecosystem, signaling that platforms can improve security posture by improving victim-engagement mechanisms rather than relying solely on transactional pattern analysis. The framework's generalizability to other sensitive domains—health, government services, cryptocurrency platforms—indicates potential for industry-wide adoption.

Looking forward, the key question centers on how adversaries adapt once they recognize that victim interviews generate actionable intelligence. Scammers may develop counter-strategies to limit victim cooperation or provide false information. Additionally, privacy considerations around recording and analyzing victim conversations require careful regulatory navigation, particularly across different jurisdictions where CASE may be deployed.

• →CASE uses conversational AI agents to extract detailed scam intelligence directly from victims, identifying patterns invisible to transaction-based monitoring alone.
• →Implementation on Google Pay India achieved 21% increase in scam enforcement volume, demonstrating tangible security benefits from victim-centric intelligence gathering.
• →The framework converts unstructured victim interviews into structured data for both automated and manual enforcement mechanisms, creating an adaptable security workflow.
• →CASE's architecture is designed for generalization across sensitive domains beyond payments, offering a replicable blueprint for other platforms facing sophisticated fraud.
• →The approach combines human-centered victim engagement with AI-powered data extraction, recognizing that social engineering intelligence requires both conversational depth and automated processing.