In the first of a two-part article, Jonathan D Hancock, head of product and innovation at The AI Corporation, looks at the impact of generative AI on fleet and mobility payment fraud.

Jonathan D Hancock, head of product and innovation at The AI Corporation

One of the most discussed technological advancements in 2025 continues to be generative AI. From its potential to revolutionise back-office and servicing operations to its increasing role in fraud management and cybersecurity, generative AI has made significant waves across numerous financial services and fintech sectors. However, a key area of focus in payments is its influence on fraud – specifically, across the fleet and mobility payment landscape.

In this fleet and mobility management context, payment fraud has long been a significant and growing concern. Fleet cards, telematics systems and digital wallets have all transformed how companies manage logistics, fuel costs and operational efficiency. Yet, this shift has opened new avenues for fraudulent activity. Fleet card fraud has long been an area vulnerable to manipulation, and with the advent of generative AI, the question arises: how will fraudsters use AI to exploit these systems, and how will we in the fraud prevention and management businesses adapt?

This two-part article series will examine the impact of generative AI on fraud within the fleet and mobility payment industry and whether that will change in the coming year. Specifically, we will assess whether generative AI has had the effect we thought it would have on the fight against fraud.

The role of generative AI in fraud management

Generative AI refers to algorithms that produce text, images, code and even videos resembling human creations. These AI systems learn from vast datasets to understand patterns and generate outputs that mimic or exceed human creativity. While generative AI’s capabilities are most visible in content creation, its application in fraud prevention and detection is increasing.

Fleet and mobility payments fraud manifests in various forms, from compromised fleet cards to fraudulent fuel transactions and even illicit use of payment systems. In the past, fraud detection systems largely relied on rule-based solutions and predictive models to detect suspicious activity. However, generative AI is revolutionising how fraud detection systems work, moving from reactive to proactive fraud strategies, learning continuously and adapting to emerging fraudulent tactics.

Did generative AI’s application in fraud detection live up to the hype?

When generative AI started making its mark, many experts predicted a transformative effect on the payments and security industry. Many expected that AI would significantly improve the accuracy and efficiency of fraud detection systems. But has generative AI delivered on these expectations?

Early successes: enhanced detection and prevention

So far, generative AI seems to have improved fraud management techniques – at least in identifying fraudulent behaviour more effectively. One key application of generative AI in fraud detection has been its ability to simulate a wide range of potential fraudulent scenarios. These simulations, created by generative models, help our fraud platforms and systems better understand what types of fraudulent activity could emerge and how to spot them in real time.

Generative AI has also enabled more sophisticated anomaly detection. Traditional systems tend to focus on predefined, historical patterns of fraudulent behaviour, which can be limited and sometimes isolated incidents resulting in high false positive rates. However, with generative AI, systems generate countless new potential scenarios, creating a more dynamic approach to fraud detection. Additionally, they can contract these scenarios against known genuine spending behaviour, which significantly impacts finding the fraudulent needles in the haystack of genuine cardholder data.

Moreover, the technology has dramatically improved authentication methods. Research from the Fleet Management Institute posits that fleet card fraud has benefited from AI-generated biometric authentication (such as voice or facial recognition) and anomaly detection in transaction patterns, making it harder for fraudsters to impersonate legitimate cardholders.

Overhyped expectations

However, while generative AI has undoubtedly made strides, it has not been a perfect solution. The primary issue has been adapting AI models quickly enough to counter evolving fraud tactics. Fraudsters continuously refine their techniques, and whilst generative AI can identify many patterns, it can also inadvertently generate false positives or miss newly emerging fraud tactics, primarily due to the incorrect dispositioning of known fraudulent transactions and accounts from which it draws its data.

Another challenge is the complexity of implementing generative AI systems in existing fraud prevention infrastructure. Fleet operators and payment providers face high costs and time required to train and maintain AI-driven fraud detection systems. So, while generative AI shows promise, it is still not (yet) a perfect solution, and human oversight is still required to complement AI-generated insights.

In the next article, we will look at generative AI and its impact on new and emerging types of fraud, along with how fleet managers can use AI to help them beat fraud in the future.