Fraud Always Follows Incentives: AI Just Brought a Map

This article is part of a special cybersecurity feature brought to you by DMM, in partnership with Opia, and written by Sara Lomax.

I’ve spent almost 15 years working in fraud prevention, and if there’s one thing I’ve learned, it’s that fraud follows incentives.

Retail promotions and rebate programs, with their promise of cashbacks, discounts and rewards, have always attracted abuse. For most of my career, that abuse was irritating rather than existential. Duplicate claims, doctored receipts, a bit of creative interpretation of the rules. Then Covid happened, everything moved online, and fraud stopped dabbling and started scaling.

And now we have a major new player on the scene: AI.

Artificial intelligence has transformed retail in genuinely positive ways. Faster processing, smoother customer journeys, better personalisation, among other benefits. But the same tools that make promotions frictionless for legitimate customers have also made them far more attractive to fraudsters. The uncomfortable truth is that AI hasn’t just increased fraud, it’s changed its shape.

Scaling Up

The scale alone should give retailers pause. According to a report from Juniper Research, cumulative global losses to online payment fraud between 2023 and 2027 are expected to exceed US $343 billion, with the US accounting for a significant share as the world’s largest e-commerce market.

In parallel, the National Retail Federation has estimated that return and promotion abuse costs US retailers tens of billions of dollars each year. Promotions don’t always sit neatly in headline fraud figures, but anyone who’s worked with rebates, loyalty offers, or high-value incentives knows how quickly losses can add up when controls are light and volumes are high.

What’s different post-Covid is speed and automation. I’ve seen AI-generated receipts that would have passed manual checks without a second glance. I’ve seen claim submissions running at volumes no human could realistically produce. I’ve seen synthetic identities stitched together from breached and scraped data that look entirely legitimate unless you know exactly what you’re looking for. This isn’t edge-case fraud; it’s industrialized.

One of the more counterintuitive risks is that AI can make fraud harder to spot, not easier. Automation improves metrics everyone likes: faster turnaround times, lower operational cost, fewer exceptions. But AI systems learn from patterns. If fraudulent behavior becomes frequent enough, it can start to look normal. At that point, the system isn’t failing; it’s doing exactly what it’s been trained to do.

I’ve seen this play out more than once. Dashboards look healthy, queues are quiet, and everyone assumes the problem is solved. Then someone steps back and looks at trends over time and realizes the system has simply become very good at approving fraud efficiently.

AI in the Driver’s Seat

Industry research backs this up. Surveys across retail and financial services show that most organizations have seen fraud losses increase year over year since 2020, with many pointing to AI-enabled and automated schemes as a key driver.

Some estimates suggest that more than half of fraud activity now involves some form of machine-generated or synthetic content, whether that’s fake documents, automated submissions or manipulated identities. The tactics might be more visible in banking, but they translate easily into retail promotions, rebates, and incentive programs.

This is where promotions become particularly exposed. They prioritize customer experience, speed, and scale. Controls are often bolted on later, if at all. Losses are written off as marketing cost until they quietly outgrow the budget that was supposed to absorb them.

I’m an accredited fraud specialist, and after nearly 15 years in this space I’ve learned to be suspicious of anything that promises instant approval, frictionless processing and zero risk. Those three things rarely coexist for long.

A Holistic Approach

The organizations that manage this risk best don’t rely on a single clever tool; they design promotions with fraud in mind. They layer controls rather than trusting one signal. They monitor behavior and trends over time, not just individual claims. And they treat AI as something that needs governance, challenge and oversight, not blind faith.

There’s a running joke in fraud circles that fraudsters used to need specialized skills. Now they mostly need a laptop, patience, and an AI tool that can generate something “convincing enough.” It’s funny because it’s uncomfortably close to the truth.

Despite the serious issues that it raises, AI isn’t the enemy; overconfidence is. Retailers that recognise how fraud has evolved, particularly in promotions and rebates, are far better positioned to protect their margins and their customers. Those that don’t may not notice the problem straight away. But eventually, the numbers stop adding up.

And fraud, as always, will already be several steps ahead.