New Visa research finds consumers are fine letting AI shop for them, but deeply unwilling to let it pay. I chat with Group President Oliver Jenkyn on why trust, not technology, is the gating factor.
VCG via Getty Images
Almost three quarters of us have used AI agents like OpenClaw or Hermes, or perhaps Meta‘s new Muse. But only 23% trust agents with their money. That’s a problem for Visa, which is looking to expand its payments empire. As Visas chief product and strategy executive writes, “Your AI just bought you a couch. Are you okay with that？”
AI agents are increasingly common. They can check Slack or summarize our email, They can help us manage our calendars, they can research information. They could also be used to track prices, re-order consumables, or even book travel, but few of us have trusted them with our wallets.
At least, so far.
Citing a survey of 2,065 US consumers, Visa says that when its name is attached to the agentic AI purchase, acceptance almost triples to 61%. That is, of course, not a statement about whether the technology works or not, or which AI agent you should use. Instead, its a statement about who consumers want standing behind them when something goes wrong, and their AI agent spends $50,000 on a designer sofa, arriving tomorrow. (Yes, I have seen a sofa that expensive.)
“AI has the potential to fundamentally reshape how people discover, buy and pay for goods and services, much like e-commerce and mobile commerce did before it,” Visa Group President Oliver Jenkyn said in the release. “While were still in the early days, trust will be foundational to driving agentic commerce adoption.”
Trust will be important.
So will the ability for our agents to do what they need to do in order to follow our instructions. So far, that‘s not off to a great start. Amazon, for instance, really does not want your AI agent surfing its site looking for bargains. It’s become a battleground that has led to lawsuits.
Discover more
Brokerages & Day Trading
Distributed & Cloud Computing
Financial
And the playing field is heating up. Visa has Trusted Agent Protocol, an Agentic Directory, an Agent Score and Visa Intelligent Commerce. Mastercard has Agent Pay. Google has AP2, an agentic payments protocol. OpenAI and Stripe have the Agentic Commerce Protocol. Every one of those is, functionally, an attempt to solve the same problem: how does a merchant know the agent knocking on its door is a real customers real agent, and not a bot with a stolen card？
The flipside of all this incredible technology is that the bad guys also have access to it for their nefarious ends, Jenkyn told me in an email interview.
Assuming the trust – and the retailer politics – get solved along with the technology, though, AI agents really do have the potential to do a lot of shopping, comparison, price-checking, and evidence-gathering for us as we make purchase decisions. And, even, make purchase decisions according to guardrails weve set up.
But there‘s probably a lot to figure out yet. Fraud detection built for humans clicking buttons doesn’t necessarily transfer to AI transacting at machine speed. Neither does dispute resolution: if your agent buys the wrong thing, or the right thing at the wrong price, the chargeback framework that protects you today isnt obviously fully applicable.
Which may be why Jenkyn thinks the first real proving ground isn‘t consumers at all. It’s procurement, invoicing and reconciliation: B2B workflows that are rules-based, permissioned and boring. And when I asked how much shopping will be autonomous in three years, he wouldnt give a number, offering instead that “human-in-the-loop” AI commerce will matter on the way to the fully agentic version.
Our full exchange, lightly edited only for length, follows.
John Koetsier: Why does trust jump from 23% to 61% when Visa is involved？
Oliver Jenkyn: “Technology often moves faster than people and businesses are prepared to embrace it. That is just human nature. The bridge between the two is trust. We saw that with e-commerce and mobile payments, and we are seeing it again with agentic commerce. In times of uncertainty and change, consumers and businesses turn to brands that they trust. Trusted brands matter more in this new digital payment world, not less. For decades, Visa has helped consumers embrace new ways to pay securely and with confidence. As AI changes the shopping experience, our role is to bring that same trust to agentic commerce.”
John Koetsier: What has to happen before consumers trust AI to buy for them？
Oliver Jenkyn: “The question is no longer whether AI can help people shop. It already is. The question is how we enable it to transact safely and securely on their behalf. Three things need to happen: consumers need to provide an authenticated, tokenized payment credential; they need to be able to personalize the agent so it reflects their preferences; and they need to set clear controls and permissions around when it can buy, how much it can spend and where it can shop. Get those three things right, and an agent can make purchases safely, securely and based on the consumers own tastes.”
John Koetsier: How much control will consumers have over what agents can spend or buy？
Oliver Jenkyn: “Consumers will stay firmly in control. The simplest way to think about it is: empower your agent to shop, but on your terms. Yes in travel, no in healthcare. Yes if the purchase is below a certain threshold, but no if it is above a certain amount. Yes if it is a merchant that I have used before, but no if it is a new seller (or at least ask me first). Consumers should be able to decide where an agent can shop, what it can buy, how much it can spend and when it needs to come back for approval. The goal is not to remove consumer control. It is to make commerce easier while enabling the agent to make decisions that genuinely reflect the consumers intention.”
John Koetsier: How do you tell a legitimate AI agent from a fraudster？
Oliver Jenkyn: “That is one of the critical challenges we need to solve. The flipside of all this incredible technology is that the bad guys also have access to it for their nefarious ends. The real question is how you spot the good agents from the bad. This isnt a new problem for Visa. Every day, we help verify identities, authenticate transactions, and manage risk across millions of payment interactions. Agentic commerce is a new channel for commerce, but it needs to be built on the same foundation of trust…transaction by transaction. Through Visa initiatives like Trusted Agent Protocol, Agentic Directory, Agent Score, and Visa Intelligent Commerce, we are working to bring the same clarity, security and trust to AI agents that consumers expect from digital payments today.”
John Koetsier: If an agent makes a bad purchase, whos responsible？
Oliver Jenkyn: “Consumers will not hand over more of the shopping journey unless the rules are clear. They need to be clear on what they approved, what protections apply, and who steps in if an agent gets it wrong. That clarity has to be designed into the experience. Payments are not simply about moving money from account A to account B. Payments require authentication, security, risk management, dispute resolution and liability protection. Through Visa initiatives like Trusted Agent Protocol, Agentic Directory, Agent Score, we are working to bring the same clarity, security and trust to AI agents that consumers expect from digital payments today.”
John Koetsier: Whats the biggest barrier today: tech, infrastructure, regulation, or trust？
Oliver Jenkyn: “Trust. Technology is moving fast, but commerce does not scale simply because something can be built. It scales when consumers and businesses trust it enough to use it again and again. Infrastructure, standards, and regulation all matter, but trust is what turns innovation into everyday behavior. As I often say, the innovation of today is the business-as-usual of tomorrow, but only if people trust it.”
John Koetsier: What are you learning from your 100+ agentic commerce partners？
Oliver Jenkyn: “The clearest learning is that partnership will win. No one player can build agentic commerce alone. AI platforms, merchants, issuers, developers, and payment networks will all have a role to play. Through offerings like Visa Intelligent Commerce, we‘re providing the infrastructure and tools that help bring those players together. But trust can’t appear only at the moment of payment. It has to be present from the first recommendation to the final confirmation.”
John Koetsier: Why do you think B2B could adopt agentic commerce first？
Oliver Jenkyn: “B2B is a very natural place for agents to prove their value because much of business commerce is already data-rich, repetitive, and rules-based. Think about procurement, invoicing, reconciliation, supplier onboarding, and payment execution. The parties are known, the permissions are clear, the data payload is rich, and controls exist. That gives agents a very natural and practical place to create value and earn trust before the experiences become more mainstream for consumers.”
John Koetsier: In three years, how much shopping will happen autonomously？
Oliver Jenkyn: “I would not put a precise number on it yet. The real impact of GenAI on commerce has not fully materialized, but the disruption, revenue opportunities and efficiency gains are coming. I have absolute conviction that AI-enabled commerce will significantly change how shopping takes place around the globe, and for the better. The measure of success will not simply be how autonomous commerce technology becomes. It will be whether commerce becomes simpler, safer, and more useful for consumers and businesses. Frankly, I expect ‘human-in-the-loop’ AI-commerce will have an important role to play in the evolution towards full agentic commerce.”
John Koetsier: Thank you for your time.