Frank on Fraud: 5 Reasons Why Artificial Intelligence Alone Is No Substitute for Fraud Expertise


Frank McKenna, Point Predictive’s Chief Fraud Strategist and resident expert on fraud prediction, recently wrote about the potential – and the limits – of AI in the fraud prevention field on his personal blog. Here are some key takeaways.

#1 – AI is a powerful tool, but it is no replacement for human knowledge

While AI tools play a supporting role, the ultimate decision-making power rests in the hands of the human fraud analyst. Recent research underscores the remarkable ability of the human brain to spot deep fakes, akin to an intuition that detects discrepancies, even if not easily articulated.

#2 – AI has an issue with false positives

AI models inherently yield false positives in the context of fraud detection. AI predictions rely on probabilities, never reaching 100% certainty. This perpetual uncertainty requires the intervention of fraud analysts to assess outputs, contact customers, and make judgment calls.

#3 – Fraud analysts can think and AI cannot

AI excels at drawing on historical data to provide insights, but it lacks the capacity for original thinking, reasoning, or adapting to entirely unfamiliar scenarios. With ever-industrious fraudsters always coming up with the next scam, analysts can adjust to new trends as they develop. In the landscape of fraud, rapid adaptation and creative problem-solving are essential attributes that AI cannot emulate.

#4 – AI needs input from experienced professionals

AI requires data to learn. And AI built for fraud sometimes requires millions of example cases to begin to recognize and react to patterns.

Those examples are called “fraud tags,” and every single one of them was identified by a human fraud analyst. In other words, AI is only as good as the fraud analyst that marked those frauds.

#5 –The Fair Credit Reporting Act can be difficult for AI to navigate

FCRA restricts specific data elements for credit assessment. Some of these, like zip codes, can provide critical insights into fraud patterns. Because fraudsters often operate within geographically clustered networks, the ability of AI models to consider these factors is paramount. Human fraud analysts remain essential to prevent AI from considering factors like zip code alone to prevent unfair rulings.

To read the original post and to keep up-to-date on Frank’s fraud insights, visit

Disclaimer: The views expressed in Frank on Fraud are the personal perspectives of Mr. McKenna and do not necessarily represent the views of Point Predictive.