Mural's CEO predicts AI agents could start 5% of cross-border payments within a year, but who pays when software errs is still unsettled.
The CEO of payments company Mural says AI software could start about 5% of international payments within the next year. That is 1 in 20. Sinclair Toffa made the forecast in a conversation with Karen Webster, CEO of PYMNTS. The coverage itself calls the number aggressive, because key questions about responsibility are still open.
Toffa said, "Agents are the new interface," comparing them to the move from desktop computers to mobile apps. AI agents are programs that take actions on your behalf, not just answer questions. He argued that new systems must be built with them in mind.
Mural is trying to standardize the software pieces that connect existing payment networks. Businesses would use one common connection point, known as an API, instead of many separate ones. Compliance records, permissions and audit trails, meaning a log of each payment, would be built in.
To work responsibly, AI agents would need their own financial accounts, tightly limited permissions and a record of every transaction. They should not get open access to a company's existing bank relationships. Authorization, liability and oversight remain unresolved, which is why the 5% forecast looks bold.
Toffa described a common problem in Latin America. A company earns dollars, converts them to local currency, then converts back to pay an overseas supplier. Each step can add exchange costs. Stablecoins, digital money meant to track the dollar, can cut some of that friction, though conversion, redemption and liquidity risks remain.
Mural works with payroll provider Deel on stablecoin payouts to workers in Latin America. Toffa noted that many people and businesses want to hold dollars but struggle to open U.S. dollar bank accounts.
The biggest open issue is who pays when the software gets a payment wrong. Webster stressed that trust must be built and liability clarified before autonomous payments can scale. Businesses that pay overseas suppliers or staff may soon be offered AI-driven options. Useful questions include:
Before letting AI send international payments, ask your provider who is liable for mistakes and whether every payment is logged and limited.
Mastercard describes AI tools that could buy things for you, and says trust is the hard part: who gets to spend your money, and under what rules.
Read more: Mastercard outlines how AI assistants could shop and pay for youWorld models, AI that learns physics from video, are being built into home robots and driverless cars, and your own videos may help train them.
Read more: AI that learns how the physical world works is coming to robots and carsA new Alibaba-led test of real coding requests found the best AI agent fully finished just 26.4 percent of tasks, so human review still matters.
Read more: AI coding helpers fully finish only about 1 in 4 real tasks
