The new American shopper can tell an artificial intelligence platform precisely what she wants, compare the answer (even while standing in the store) and abandon the merchant at checkout because the wrong payment option appears.
That combination of sophistication and impatience is changing the merchant-consumer equation. Consumers have acquired better tools for finding products, comparing offers and deciding how to buy. Merchants, meanwhile, still have to answer some stubbornly basic questions: Do we recognize this customer? Is this AI agent legit? Can it read the catalog? Can the agent assess and apply promotion and loyalty offers? And can she pay the way she expects?
The answers won’t surprise any merchant who puts the consumer at the center of their business. Michele Herron, senior vice president and head of North America Value Added Services at Visa, told PYMNTS CEO Karen Webster that AI adoption is showing up among precisely the consumers merchants most want to understand.
“The data is telling us that the best shoppers are AI enabled,” Herron said. People who browse more often and check their phones while shopping are valuable consumers who also show a strong affinity for AI.
That changes the merchant calculus around AI. Early assumptions cast shopping agents as useful primarily for consumers who wanted software to handle an unwanted chore. Herron noted that joint research between Visa Acceptance and PYMNTS Intelligence to produce the sixth annual Global Digital Shopping Index indicate that AI and people who love to shop are highly correlated.
Their shopping is also hard to contain within conventional measures of visits or channels. Webster pointed to 52 buying days a month among AI users versus 21 for nonusers, reflecting multiple shopping occasions within individual calendar days. High-income consumers record 54 digital buying days. A digital buying day is defined as an instance over the course of a given month in which a consumer uses a digital device to shop.
The phone joins those occasions together. Webster described the “always-on shopper” as one for whom “the phone is literally the store,” with browsing and buying blending rather than occurring as distinct activities.
What merchants can’t assume is that being present on that phone is enough.
Payment choice has become part of merchant selection. Webster said the study data finds that roughly two-thirds of consumers consider payment preference and acceptance when deciding where to shop, compared with 58% two years ago.
Herron sees a tension there. Payments technology has spent years trying to remove itself from the consumer’s attention, yet consumers still care deeply about what happens at the moment of payment.
“Payments are best when they’re invisible, but they have to work exactly the way that consumer wants to,” Herron said.
Merchants Need to Know the Shopper Again
AI makes the customer-recognition problem more difficult because an agent may arrive carrying detailed instructions from a shopper without giving the merchant much insight into who that shopper is.
That is potentially valuable information. Webster said an AI prompt can provide far richer context than a traditional keyword search, describing what someone wants, the circumstances surrounding the purchase and the attributes that matter.
Yet Herron said merchants frequently can’t connect that intent to an existing customer.
“They’re not even really identifying that it’s ‘Karen behind the agent,’” she said, adding, “Merchants don’t want to lose the loyalty experience. They’ve got CRMs that are built. They’re trying to track their consumer at every approach, and that’s missing right now.”
Before merchants can extract much value from that information, however, many have an even more rudimentary job to complete. The study finds that only about 15% of merchants have machine-readable catalogs. If agents can’t reliably interpret products, prices and inventory, they can’t efficiently deliver consumers to a transaction.
Herron said an AI agent forced to navigate a website designed primarily for people can encounter “about a 70% failure somewhere in that purchasing pathway.”
Visa is working with clients on making product information machine-readable and helping maintain and enrich those catalogs, she said. The network is also working on a related identification problem. Separating legitimate shopping agents from automated traffic merchants have learned to regard as dangerous. Trusted-agent protocols can help establish that an agent has been verified rather than treating every bot-like visitor as suspect.
Tokenization provides another piece. Herron said Visa is looking at how tokenization, authentication, payment instructions and transaction signals can help identify legitimate agents and reconnect their activity with the consumer behind them.
For merchants, the payoff is fairly direct. Better product data can make inventory discoverable. Better agent identification can prevent valid shoppers from being rejected. Better customer recognition can preserve loyalty and personalization. Payment choice can keep an otherwise successful shopping journey from ending elsewhere.
Consumers get something equally concrete: AI assistance without being forced to give up the controls and protections they already expect when they pay.
The industry doesn’t need to predict exactly which AI platform or agent will dominate before beginning that work. Herron’s advice was to make the underlying commerce infrastructure ready for whichever one consumers choose.
“Embracing the protocols, starting to embed some of the secure experiences that we already have available, will set you on the path to success,” she told Webster.
Watch the full interview with Michele Herron to learn more about:
- Why consumers remain reluctant to give AI complete authority over a $300 purchase, even when they are comfortable delegating other shopping tasks.
- What spending limits, category controls and remediation options may be needed before consumers trust agents to complete purchases for them.
- Why Herron believes the merchant will remain central to agentic commerce rather than being reduced to a passive supplier behind a handful of AI platforms.
