Trustmi says USD $5 billion in payment errors detected

Trustmi has published research showing that $1 in every $48 of enterprise payment value involved an error it prevented. The analysis covered USD $240 billion in payments over one year.

Those findings equate to USD $5 billion in payment errors detected before funds were released. Duplicate invoices and duplicate payments accounted for about half of the affected payment value. Data-entry mistakes and currency mismatches each represented 20%, while the rest involved payments to the wrong business entity or missing processing steps.

The figures point to persistent weaknesses in accounts payable controls, particularly in large organisations that already use purchase order matching, approval processes and duplicate-detection checks within enterprise resource planning systems.

Trustmi also conducted a separate retrospective review of one year of ERP records at a global enterprise operating across 20 business entities in 15 countries. It identified about 85 probable duplicate invoice cases, including 59 in which both copies were paid, with roughly USD $560,000 in flagged second-payment value.

The analysis found that small differences in invoice records often allowed duplicate payments to slip through. In 61% of cases where an invoice was paid twice, the invoice numbers were not identical. Added letters, formatting differences or revisions made probable duplicates appear to be separate transactions.

Timing was another factor. Some 58% of duplicates that were paid twice were entered into the ERP system within a week of the original invoice, and more than half cleared in the same payment run.

Where errors emerge

The review also suggested that standard workflows failed to stop many duplicate payments. Eight in 10 of the duplicates that were paid twice had gone through the standard purchase order-matched invoice process, indicating that matching and approval steps alone were not enough to identify all likely repeat transactions.

Patterns also emerged across the calendar and working week. Duplicate payments peaked in December and January, with another rise in July. Second copies were entered most often on Fridays and Mondays and least often on Wednesdays.

These trends suggest that routine pressure points in finance operations may create openings for avoidable mistakes. Year-end periods, new reporting cycles, and the start and end of the working week are often associated with higher processing volumes and tighter deadlines.

Trustmi said the limits of conventional controls stem from the way many finance systems validate transactions only within their own records. Without access to broader payment history across multiple systems, a duplicate invoice that differs slightly from an original may not be flagged before payment is made.

It argued that a wider historical view of supplier and payment activity is needed to identify anomalies across invoices, transaction amounts, payment behaviour and vendor details.

"Despite long-established payment processes, checks and controls, there's a clear pattern of significant payment errors that continue to fly under the radar," said Eli Ben Nun, Co-founder and CTO at Trustmi.

"More disparate checks aren't the answer, but errors don't have to be an inevitable cost of business. Using behavioral AI, it's possible to spot fraudulent payments and those made in error before the money moves," Ben Nun said.

Broader context

The research adds to a wider debate in corporate finance over whether established accounts payable systems are equipped to detect increasingly subtle forms of duplication and error. Many organisations rely on rule-based checks within ERP platforms, but those checks can struggle when invoice details vary only slightly or when records are spread across business units and jurisdictions.

Trustmi said its analysis excluded recurring monthly charges and that the duplicate-payment review identified probable duplicates rather than confirmed losses. The flagged amounts also do not establish whether payments were later recovered.

The company, founded in 2021 and based in New York City, focuses on payment security and fraud detection. It says it protects more than USD $240 billion in payments annually.

The ERP review covered records from 20 business entities in 15 countries and identified probable duplicate invoices through similarities in vendor information, payment amounts and invoice numbers.