The archaic interface of Citibank’s loan system required checking multiple boxes to do an interest-only payment – a confusing design that proved very costly.

Our first tale is a banking blunder so epic a judge called it “one of the biggest blunders in banking history.” In 2020, Citibank meant to wire $7.8 million in interest to lenders of Revlon. Instead, thanks to a confusing interface, they accidentally wired nearly $900 million – paying off the entire loan by mistake. Three people approved the transfer, yet none realized that the software (a system called Flexcube) defaulted to sending the full principal unless you checked not one but three separate override boxes. They missed two of the boxes, and off went almost a billion bucks. Citibank scrambled to undo the error, but some lenders refused to return the money. A court later ruled Citibank had to eat a $500 million loss as a result.

UAT Fallout: How did such an obvious mistake get through UAT? The interface was so unintuitive it fooled even seasoned staff. It seems no one tested the realistic test case or user scenario of “operator attempts an interest-only payment.” The design assumed users would read the manual (hint: they didn’t). Poor UAT meant no one experienced the product from the end-user’s confused perspective before go-live. In short, the software worked as built, but not as any sane user would expect – a classic UAT failure.

Quell to the Rescue: Quell’s AI + human-based testing would have caught this UI nightmare before it cost real money. Imagine a Quell “operations” agent stepping through the payment process from an end-user perspective. It would try to make an interest payment and flag the multi-checkbox booby trap. With Quell’s multi-disciplinary approach, one agent could act as a forgetful user who checks only one box, while another acts as a compliance officer verifying totals. The discrepancy – “Why are we sending $900M instead of $7M?” – would have been caught in UAT. Quell’s intelligent agents don’t assume the user is perfect, and they don’t skip the fine print. This horror story would have been a non-event with Quell: just another bug caught in the test environment instead of front-page news.