Cybercrime is expected to cost the world $10.5trillion annually by 2025 but machine learning is helping merchants and financial institutions to fight back against criminal activity.
Monica Eaton, CEO of Chargebacks911, an international chargeback management and prevention company which provides SaaS solutions for managing chargebacks, discusses why artificial intelligence (AI) has always been at the forefront against fraud.
The emergence of generative artificial intelligence has attracted lots of excitement over the years, but while many companies behind the rise of AI applications have seen their valuations skyrocket, the technology is not unfamiliar territory for the finance sector—specifically in the chargeback space.
Machine-learning (ML) solutions were deployed many years ago to aggregate and segment large sets of transaction data to help guide policies, operations and decision making for banks and businesses.
This technology is especially critical today, where it is nearly impossible to counter online fraud and chargeback abuse manually, especially with cybercrime as a whole expected to cost the world $10.5trillion annually by 2025.
With everyone talking about AI and its overall potential, I’ll aim to answer what it is, what it can do, and what it has been doing for many years to keep stakeholders safe.
A close up of AI
As portrayed in the movies, AI is simply a virtual being with intelligence comparable to a human. This emerging technology is being trusted enough to be conversed with, asked questions and solve problems in real time without any human oversight.
However, what OpenAI, Google and others have created is far different. ChatGPT can only complete specific tasks based solely on the information on which it’s built, whereas a human brain would undertake tasks with distinct perspectives, opinions or personalities.
Large language models (LLM) like ChatGPT can draft an unlimited amount of accurate and well-written content, similar to how autocorrect works on your phone. By learning what kind of words follow certain questions, and by accurately predicting their answers, LLMs can convincingly present themselves as living, responsive beings. However, this can fall short when it doesn’t understand the meaning or is working on the limited context behind any of these words or questions.
With a large enough dataset and enough tweaking by its human programmers, LLMs can still be very realistic and produce seemingly human interactions, but programmers and users need to be cautious that AI tools could cause mistakes, disruptions, or misguidance if the information which responses are based on are inaccurate or outdated.
Using AI to combat fraud and reduce chargebacks
Since AI can be prone to error, how should we mitigate risks when using it to fight fraud? While we must ensure that AI tools are working within the right perimeters and are accurate and up to date, AI (or more accurately, ML) in anti-fraud applications have become adept over time at finding fraud and representing chargebacks.
The anti-fraud industry can quickly spot irregularities and patterns within data, something that computers are uniquely good at. For example, if every field in an order form is filled in instantly, instead of taking a little time as most humans do, this could indicate that the form is being filled in automatically rather than by a person, a telltale sign of fraudulent activity. Another example would be AI automatically flagging a transaction for inquiry if the distance between shipping and billing address is drastic.
ML can also effectively spot irregularities in chargeback management, even if a person has simply issued chargeback claims too frequently. Completing tasks on a per-retailer basis is also crucial, so the machine-learning algorithm learns the specific nuances of how fraudulent chargebacks affect a particular merchant’s business. Signs of chargebacks (both valid and invalid) can be learned at an expedited rate with faster connections than humans—contributing to a higher customer satisfaction as it only lets through genuine transactions in an efficient manner.
A trusted and mature technology for retail and fraud prevention
When using AI to prevent fraud and chargebacks, there are certainly going to be trials, errors and learning opportunities along the way, but we’re seeing the technology become more mature as retailers around the world can put their trust in it and provide it with more reliable data on which to base its decisioning. If we want to move forward successfully with AI, we have to be realistic about its capabilities over the coming years, as more retailers implement it into their workflows.
