How AI Has Been on the Front Line Against Retail Fraud

Over the last few years, there has been great momentum in the discussion about the current and potential impact of generative artificial intelligence (AI). Much like streaming providers during the pandemic or shovel salesmen during the gold rush, companies that supply the power behind AI applications have seen their valuations skyrocket. However, in the finance sector (and more specifically, the chargeback space), AI isn't a new concept.

Our industry deployed machine learning (ML) solutions many years ago to aggregate and segment large sets of transaction data to help guide policies and decision making. And with cybercrime as a whole expected to cost the world $10.5 trillion annually by 2025, it's nearly impossible to counter the issue manually. With that, let’s explore what AI is, what it can do, and what it has been doing for many years to keep retailers safe.

AI Under the Microscope

AI in its truest form (or what some may have seen in the movies) is a virtual being with intelligence comparable to a human who, like humans, could be conversed with or depended on to solve problems in real time without human oversight. But this isn't what OpenAI, Google and others have created. For example, ChatGPT can perform particular tasks in particular ways, but unlike human brains, it cannot learn new tasks, nor does it have distinct perspectives, opinions or personalities.

A large language model (LLM), like ChatGPT, trawls as many pieces of written content as possible to build a model of what kind of words go with other words, much in the same way that the autocorrect in your phone operates. It learns what kind of words follow certain questions, giving it the ability to answer questions in a realistic way, as if it were a thinking being. It doesn’t understand the meaning or context of any of these words, but given a large enough dataset and enough tweaking by its human programmers, a LLM can be very realistic. However, while AI applications like LLMs can produce seemingly human interactions, if the information which responses are based on are inaccurate or outdated, then AI tools could cause mistakes, disruptions or misguidance.

AI in Fraud and Chargebacks

So, if AI can be limited and potentially prone to error, then should there be restrictions in how it’s used in fighting fraud? The answer is no, but those utilizing AI to identify and eliminate fraud should ensure the parameters their tools are working off of are accurate and up-to-date. For some time, AI (or more accurately, ML) in anti-fraud applications has become adept at finding fraud and representing chargebacks. This is because the anti-fraud industry uses it for something that computers are uniquely good at: spotting irregularities and patterns within data. For example, if every field in an order form is filled in instantly, instead of taking a little time as a human being fills it in, this could indicate that the form is being filled in automatically rather than by a human being, a telltale sign of fraudulent activity. Another example is if the distance between the shipping and billing addresses are drastic, this can automatically flag a transaction for further inquiry.

ML can also look for patterns in chargeback management, which can be as basic as whether a person has repeatedly issued chargeback claims. More importantly, this can be done on a per-retailer basis, so the machine-learning algorithm learns the specific nuances of how fraudulent chargebacks affect a particular merchant’s business. Because it's far faster than a human operator at learning these signs of chargebacks — both valid and invalid — and can make connections that people just couldn’t make as quickly, it contributes to customer satisfaction by reducing the number of false positives and only letting through genuine transactions in an efficient manner.

A Mature and Trusted Technology for Retail

Being realistic about the capabilities of AI is going to be crucial for many over the coming years, as more retailers bring various forms of AI into their workflows. Although there are going to be some trials, errors and learning opportunities along the way, the use of AI and ML to prevent fraud and chargebacks is a mature technology that retailers around the world can trust if in experienced hands.

Monica Eaton is the founder and CEO of Chargebacks911 and Fi911, as well as chief information officer of Global Risk Technologies. Chargebacks911 is the first global company fully dedicated to remediating chargebacks and helping to eliminate first-party fraud and misuse. Fi911’s pioneering DisputeLab™ tool streamlines chargeback management for acquirers, automating legacy processes and standardizing methods that simplify and speed the end-to-end workflow, improving the customer experience and accountability for all stakeholders.