How are LLMs impacting the fraud landscape?

Large language models (LLMs) and AI agents are not just buzzwords—they're reshaping how we think about fraud detection and prevention. In this episode, we’re joined by  Frank McKenna, a well-known fraud expert and Chief Fraud Strategist at Point Predictive (as well as the voice behind Frank on Fraud, a widely respected blog in the fraud space) and Inscribe CEO Ronan Burke. 

This discussion dives deep into the transformative potential of these technologies, their practical applications, and how they empower fraud analysts to stay ahead in an ever-evolving landscape. We also share practical tips for integrating AI into risk management workflows while addressing the challenges and opportunities it presents for the financial services industry.

Evolution of Fraud Detection

• Fraud prevention has advanced significantly over the decades, moving from manual processes in the 1990s to rules-based systems, big data analytics, and now AI-driven methods.
• Frank highlighted that in the early days of fraud detection, tools were minimal, relying on rudimentary ranking systems to flag suspicious activity. The introduction of machine learning in the 1990s, such as Falcon, marked a significant leap.
• Today, technology leverages advanced capabilities like biometrics, device profiling, and large language models (LLMs) to address increasingly complex fraud patterns.

“When I started in fraud back in the 1990s, there were only two types of fraud: credit card fraud and check fraud. Now, the landscape is so much more complex, but the technology has evolved just as dramatically,” Frank explained. 

The role of AI in modern fraud detection

• LLMs like GPT and Claude are enabling fraud detection teams to analyze and synthesize vast amounts of data quickly and accurately.
• Network effects amplify fraud detection by pooling insights across ecosystems, while LLMs provide the ability to plan and perform actions autonomously, which is transformative for the industry.

“Network effects allow us to detect patterns across ecosystems, and LLMs take it further by synthesizing data and even planning actions—a completely new capability for the fraud industry,” Ronan shared. 

Real-world applications of LLMs

• Frank demonstrated Claude’s new analytics feature, showcasing its ability to analyze data, generate visuals, and produce comprehensive summaries within seconds. This capability significantly reduces the time and effort required for data-driven decision-making.
• Ronan shared how Inscribe's AI fraud analyst uses LLMs to corroborate information across diverse data points—for example, flagging risks when a submitted pay stub originates from a business that has closed.

“With Claude, you can load a dataset, prompt it with a simple question, and within 60 seconds, it delivers analysis and visuals that would have taken a human days to produce,” Frank said.

AI Agents: The next frontier

• Frank described AI agents as “supercharged macros” capable of automating specific, repetitive tasks and acting as intelligent assistants for fraud analysts. He emphasized their potential to revolutionize workflows by enabling individuals to manage teams of virtual experts.

“Think of AI agents as PhD-level experts who can do tasks for you. The possibilities are endless,” Frank said. 

• Ronan elaborated on Inscribe’s approach to AI agents, detailing how their system provides natural language explanations for fraud findings, corroborates diverse data sources, and performs advanced web research to surface critical insights.

“Our agentic systems at Inscribe can perform mundane, repetitive work exceptionally well—from application reviews to web research—freeing up fraud analysts to focus on higher-order tasks,” Ronan shared. 

Will AI Agents replace frontline fraud analysts? 

• While AI tools are transforming fraud detection, both Frank and Ronan agreed that human fraud analysts will remain essential. These technologies will augment analysts’ roles, enabling them to focus on higher-order tasks and strategic decision-making.
• Ronan noted the potential for AI adoption to create a divide between resource-rich institutions and smaller financial organizations. Ensuring accessibility for all institutions is crucial to maintaining a fair and competitive landscape.

“AI won’t replace fraud analysts; it will enhance them,” Frank said. “Embrace it, learn it, and use it to excel.”

• Ronan noted the potential for AI adoption to create a divide between resource-rich institutions and smaller financial organizations. Ensuring accessibility for all institutions is crucial to maintaining a fair and competitive landscape.

Practical tips for getting started with AI in 2025

1. Education: Subscribe to newsletters and resources on AI to stay informed about advancements and use cases.
2. Experimentation: Start with small, manageable projects to explore AI’s capabilities—whether creating meeting agendas or analyzing historical data.
3. Adoption: Invest in AI tools and provide training for team members to understand and integrate these technologies effectively into their workflows.

The episode concludes with optimism about the future of AI and its potential to empower fraud analysts, enhance decision-making, and drive efficiency. The speakers emphasized the importance of embracing AI as a collaborative tool rather than a replacement for human expertise.

Listen to the full episode on your favorite podcast platform to dive deeper into these insights and hear bloopers at the end!

Sources cited 

• Frank on Fraud Blog
• Introducing the analysis tool in Claud.ai
• Claude.AI New Analysis Tool Is Fraud Fighter Power-Up