tl;dr
Fraud detection is the use of AI and rules-based systems to identify suspicious patterns in financial transactions, a duplicate invoice, an unusual payment destination, an amount just under an approval threshold, before a payment goes out.
Because fraud typically shows up as a deviation from a normal pattern, it's a strong fit for machine-learning models trained on what normal activity looks like for a given vendor or account, flagging what doesn't fit rather than relying on a fixed list of rules that fraud can be designed around.
Back to the glossary