Poor Financial Data Threatens AI Rollout in Payments

Payments companies generate vast amounts of data from cards, digital wallets, buy now, pay later products and cross-border transactions, but legacy finance systems often retain only batched or aggregated records. Ben Catterall, global head of solutions engineering at Fynapse, said AI cannot make reliable decisions without the context behind each transaction, raising the risk that flawed inputs will magnify errors in reconciliation, pricing, forecasting and financial control.

PYMNTS reported on July 23, 2026, that poor data quality prevents 46% of AI proofs of concept from reaching production. At one multinational payments client operating across nearly 18 countries, transaction-level analysis uncovered average foreign-exchange spreads of about 2% on roughly $100 million in cross-border payment volume, implying as much as $2 million in annual lost revenue. Fynapse client T-Mobile, meanwhile, can process 200 million journal lines an hour in real time.