Shoppers will let AI browse, but they still want their hand on the wallet | Retail Customer Experience
Agentic commerce gains ground fastest wherever the financial risk is small, the decision is familiar and the shopper can see and control the guardrails around it.
Nuvei’s latest research into U.S. shopper attitudes toward AI-assisted commerce points to a familiar pattern in retail technology: people adopt the convenience long before they hand over control. The headline is clear: 58% of consumers already find AI useful for researching products and comparing prices, and 39% say they’re using it more while shopping than they were just three months ago.
But ask that same group whether they’d let an AI agent spend their money without checking in first, and the enthusiasm evaporates. 56% say they would never allow AI to spend money without their approval, and just 1% want the shopping process fully automated, start to finish.
That’s a wide gap between two closely related behaviors. Put simply, shoppers have effectively deputized AI as a research assistant. They haven’t deputized it as a wallet.
Checkout is where trust starts to dissipate
When researchers asked people what specifically worries them about an AI agent buying on their behalf, the top answer wasn’t a bad product match. Seventy six percent of consumers named fraud or unauthorized charges as their biggest concern, more than double the 33% who worried about the AI buying the wrong item.
The key takeaway: consumers aren’t questioning the AI’s judgment nearly as much as they’re questioning what happens to their money once a decision gets made.
Reversibility ranks just as high. 84% of consumers say the ability to cancel or reverse an AI-initiated purchase after checkout is extremely or very important to them, with 64% calling it extremely important. Shoppers aren’t asking for a smarter agent. They’re asking for a safety net, since they haven’t technically completed the purchase themselves: a charge they recognize, a limit they set in advance, and an undo button if something goes wrong.
Consumers are already describing the guardrails they want
Perhaps the most useful finding is how specific consumers get when asked what would actually make them comfortable letting AI buy on their behalf. 61% pointed to spending limits and 59% pointed to a required human approval checkpoint before checkout.
That’s close to a specification written by the customer base itself. For businesses, the fix is about creating visible and understandable control that shoppers can see and set for themselves such as spend caps, category limits, and a checkpoint before money moves.
A trust curve retail has seen before
None of this should feel unfamiliar to anyone who’s watched a new payment technology arrive in stores or online before. Contactless cards and tap-to-pay went through nearly the same arc: initial curiosity, a long stretch of people using the convenience but still wanting a receipt and a visible confirmation, and only gradually a willingness to trust the tap without a second thought. Mobile wallets followed a similar path. They were widely adopted for storing loyalty cards and offers years before shoppers trusted them at checkout, until issuers built recognizable, visible security cues like tokenization and biometric confirmation into the moment of payment itself.
The lesson from those rollouts is that trust in a new payment mechanism doesn’t arrive because the underlying technology becomes more secure. Cardholders were never in a position to verify that themselves. It arrives when the interface gives people a visible sense of control: a prompt, a limit, a confirmation, something to say “yes, this was me.” Agentic commerce appears to be following the same script, just compressed into a much shorter timeline because AI adoption itself is moving faster than any prior payment technology cycle.
Where consumers still want the purchase to happen
Fifty eight percent of consumers say they would still prefer to complete an AI-assisted purchase on the retailer or brand’s own site or app, compared with just 12% who would be comfortable completing it inside the AI platform itself. 62% say they’re not comfortable letting AI complete a purchase without a traditional checkout page at all.
That’s a clear signal about where the transaction is expected to land. It suggests shoppers see the AI as a shopping companion, something that helps them think, compare, and shortlist, while still wanting the brand they trust to be the one that closes the sale. For retailers, this is extremely encouraging. Now, businesses building their own AI-assisted shopping experiences on their own digital properties are aligned with where customers want to land, rather than inside a third-party agent’s ecosystem.
What this means going forward
Three things stand out from the data for any business weighing where to invest as agentic commerce matures.
- Visible controls beat invisible security. Consumers have told researchers, in their own words, what would earn their trust: spending limits (61%) and an approval checkpoint (59%). Businesses building AI shopping experiences should design those controls into the front of the experience, not bury them in settings menus.
- Reversibility has to be built in, not bolted on. With 84% of consumers rating the ability to undo an AI-initiated purchase as important, an easy, visible path to cancel or reverse a transaction should be treated as core to the experience, the same way free returns became a baseline expectation in e-commerce.
- Own the checkout moment. With 58% of consumers wanting the purchase to close on the retailer’s own site or app, and only 12% comfortable finishing inside the AI platform, businesses have a reason to build first-party AI shopping tools on ground they already control, rather than waiting for third-party agent ecosystems to mature and ceding that moment, and the customer relationship attached to it, to someone else.
Where agentic commerce is likely to gain traction first
Not every category of shopping carries the same trust burden, and the research points to where agentic commerce is likely to take hold earliest: purchases that are repetitive, predictable, and already low-stakes for the shopper.
Consumer packaged goods and household replenishment are a natural starting point. Reordering the same laundry detergent, coffee, or paper towels involves little of the “did it pick the right item” risk that worries consumers, since the product and price point are already known quantities.
Subscriptions and recurring services are another likely early mover. Shoppers already trust a recurring charge to a service they’ve signed up for, and agentic commerce mostly automates decisions, like swapping a plan tier or adjusting a delivery date, that are incremental rather than a fresh purchase.
Food and beverage, particularly grocery reorders and restaurant reordering, sits in similar territory. These are frequent, lower-dollar transactions where a wrong call is easy to correct and low-cost to the shopper, which lowers the bar for letting an agent handle it.
Travel and booking categories may follow once trust in reversibility catches up, given how often changes and cancellations already happen in that space, but the higher price points involved likely mean it lags the more routine categories above.
Across all of these, the common thread is the same one running through the survey data: agentic commerce gains ground fastest wherever the financial risk is small, the decision is familiar, and the shopper can see and control the guardrails around it.
Retail – GeneralSelf-CheckoutAIConsumer BehaviorMarketing
Related Media
Blog
AI is accelerating commerce, but service-led AI is where brands win
Blog
The now and next steps retailers should take in the agentic commerce era
Article
Consumers wary about AI shopping assistants but Amazon is changing perceptions
Blog
When AI becomes the shopper: Why retailers must rethink customer experience
Article
Retail AI 2026 predictions: Retailers, consumers driving big growth
Blog
Beyond the hype: What retailers actually learn when agentic AI hits the sales floor
