Elavon's CEO says banks should go slow on AI shopping agents to avoid repeating 25 years of fraud mistakes, and he says the bank-backed Paze wallet is ready to pilot.
The head of a major payments company is urging the industry not to rush AI shopping agents. Wally Mlynarski, CEO of Elavon, a U.S. Bank subsidiary, says the bank-backed mobile wallet Paze is "ready to do pilots." He also warns that moving too fast could harm consumers. He spoke at the Money20/20 conference in Las Vegas last week.
Elavon is an Atlanta-based company that helps merchants take payments. Mlynarski says the industry has spent 25 years chasing online fraud since e-commerce began, and "we've made all the mistakes." He pointed to card security codes, industry data security standards and keyed-in card transactions. All were meant to manage risk.
"We could go screw that up again with agentic commerce, or we can actually focus on safety and soundness," he said. Agentic commerce means AI assistants that shop and pay on your behalf.
He is not alone in his caution. The article notes that JPMorgan Chase's Jamie Dimon and PNC's Bill Demchak have raised liability concerns about AI agents acting for customers. In plain terms, the open question is who covers the cost if an agent buys the wrong thing.
Mlynarski said some recent launches are "not necessarily thinking about all the safety and soundness." Meta's personal AI agent Muse now uses Stripe's Link wallet for payments. Consumers have also started using Instinct and OpenAI's dots.
Paze is a mobile wallet from Early Warning Services, which is owned by seven large U.S. banks, including U.S. Bank. Mlynarski helped shape it during his time at Bank of America. He says it can work within agentic commerce and address agent hallucination, which is when an AI makes things up or gets facts wrong.
He also expects a fight over who gets paid. The mobile wallet wars of the 2010s, with Apple Pay, Google Pay, PayPal and later Paze, were about making the economics work. That proved difficult. Adding agents to the chain of consumer, merchant, acquirer, issuing bank and card networks, he said, "is going to be a challenge."
The article gives no pilot dates, partners or details on how a Paze pilot would handle mistakes by an agent. Mlynarski said standards and protocols for agentic technology are still being sorted out. His pitch is patience: "If we race into it and do it incorrectly, consumers are going to be harmed."
Before letting an AI assistant buy for you or your business, ask who pays when it gets a purchase wrong.
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
