The accelerations in eCommerce Delivery, Marketplaces and the gig economy have companies scrambling to make better decisions, faster. Model-based applications are being applied to address wider areas of fraud, as they can detect threats more precisely and efficiently.
Within the past year, Artificial Intelligence has made major impacts in Card Not Present and Card Present channels. Industries that showed significant growth before the pandemic experienced an even higher volume of expansion after lockdowns had been implemented. This volume growth also introduced a growing number of fraud trends on emerging eCommerce platforms.
The gig industry has been a new example of this. As multiple platforms are created in response to the pandemic to connect customers, businesses, restaurant services and delivery drivers, a growing fraud trend was to register several false accounts and publish fake marketplace listings. Traditional methods of preventing fraud, such as rules-based systems, failed in this category to keep up with the use of common text in these posts, how they were categorized and how fast new fraud schemes developed. Rules-based systems with a supervised learning structure are only limited to identifying threats that fall within their known rule structure or within labeled events. In such cases, an unsupervised AI model proved to be important in fighting fraud, as it can adapt to detect new threats without knowing what they are.
AI is also able to recognize emerging patterns more quickly than a risk analyst. Risk analysts should still review patterns AI or ML uncovers and how this impacts risk scoring decisions, but this supervised machine learning approach allows organizations to pivot or adjust risk strategies as quick as possible. Organizations and analysts tend to focus anti-fraud efforts on transaction events where fraud has a direct financial impact, but fraud occurs throughout the account creation and transaction lifecycle. With a growing number of volumes in transactional communication and a distributed workforce, human review of these actions across account creation, login and other events is simply not efficient or scalable. One could argue that an automated solution can be implemented in any sector and provide an increase in ROI over time, but the real question is how rules-based systems play a role and for how long?
For More Information: