The age of AI shopping is here. Algorithms can now decide, transact, and deliver. And buyers across the world are being asked a new question: how much of their shopping are they truly willing to delegate to machines? Recent surveys by Capgemini Research Institute, Bain & Company, and Checkout.com reveal a telling tension between curiosity and control. Global consumers are intrigued by the idea of AI doing the heavy lifting in their shopping lives. But, at the same time, they are deeply hesitant to hand over full command.
The promise: Consumers want AI in their shopping journeys
In Capgemini’s late-2024 study spanning 12 countries across North America, Europe, and Asia-Pacific, 71% of consumers said they want generative AI integrated into their retail experiences. This enthusiasm extends well beyond novelty. More than half of respondents said they now use AI tools to help them search, compare, or even design products. In some markets, consumers are turning to generative AI for shopping inspiration as often as they use search engines.
That number marks a global shift in shopping behavior. It suggests that consumers expect AI to be woven into their daily interactions. Many already rely on it for recommendations, styling advice, even virtual try-ons. AI is becoming a default part of the discovery process.
But what consumers are embracing, for now, is AI as an assistant, a co-pilot in browsing, not a substitute for human judgment. The report also found that satisfaction with generative AI experiences has in fact declined slightly year on year. While the appetite is strong, the execution hasn’t yet caught up to the promise.
The threshold: From assistance to autonomy
That threshold—between assistance and autonomy—is precisely where agentic AI enters the conversation. Bain & Company’s 2025 survey of American consumers explored what happens when AI stops recommending and starts acting. They found that while 72% of U.S. consumers have used AI in some form, only 24% said they would feel comfortable letting an AI agent make a purchase on their behalf.
It’s a striking contrast: nearly three-quarters of people engage with AI, but barely a quarter are ready to hand it a credit card. This “trust gap” lies at the heart of AI’s next evolution. And yet, despite all this disparity, ChatGPT has already rolled out instant checkout in the U.S., testing how ready people really are to let AI complete a purchase on their behalf.
The hesitation is rooted in a mix of psychology and practicality. People worry about errors and fraud. They wonder who is accountable if an AI buys the wrong size, overspends, or mishandles their data. Bain’s analysis shows that trust only begins to build when consumers retain veto power, the ability to review, approve, or cancel transactions before they happen. In other words, AI may be autonomous, but consumers still want a kill switch.
The everyday test: Where consumers draw the line
A Checkout.com-backed survey in the UK, cited by Fashion Network, offers a glimpse into how consumers are sorting their comfort levels by purchase type. Four in ten UK consumers said they’d be happy to let AI handle routine, low-emotion purchases, like groceries, toiletries, or household staples. On average, they would trust an AI agent to spend up to £200.
However, the same study found that 40% of respondents said letting AI choose or buy gifts for loved ones would feel “too impersonal.” Consumers seem to be building mental walls between functional shopping (which they are willing to automate) and emotional or identity-based shopping (which they still want to control).
Younger shoppers, though, are already pushing those boundaries. Among 25- to 34-year-olds, 70% said they’d be comfortable allowing AI to make purchases. Nearly half of that age group already use AI for product research or price comparisons. For digital-native consumers, AI-mediated commerce feels efficient and sensible.
The question: Where does trust end?
Underneath these numbers sits a profound behavioral question: what does shopping mean to the modern consumer? For decades, consumption has been an act of self-expression; what we buy tells others who we are. Handing that over to a machine, even partially, can feel like losing part of that identity.
At the same time, the appeal of frictionless convenience is undeniable. In simple terms, people want to automate the mundane so they can curate the meaningful.
This duality explains why the adoption curve for agentic AI may not be linear. Consumers will likely embrace it first for repetitive, low-stakes categories (think laundry detergent or pet food) and resist it for fashion, gifting, or luxury, where emotion and personal taste remain non-negotiable.
Across all three surveys, one theme rises above the rest: trust. 40% of UK respondents worry about fraud or data misuse if AI agents start transacting for them. In Bain’s U.S. study, the same anxiety surfaced. Trust in brands and payment systems determines readiness for automation. What’s fascinating is that this tension is universal. Whether in the U.S., Europe, or Asia, the same pattern emerges, optimism mixed with reservation. As Capgemini found, shoppers are less satisfied with how AI is being implemented than they were a year ago. Despite cultural and economic differences, global consumers everywhere are negotiating the same trade-off: convenience versus control.
Ultimately, trust will determine how quickly AI shopping scales. Consumers want transparency. They want to know who is in control, what happens if something goes wrong, and how their data is used.
The takeaway: The human limit of machine shopping
AI may soon be capable of managing entire shopping journeys, predicting needs, placing orders, or optimizing prices. Yet these surveys make one thing clear: global consumers are setting boundaries.
They are eager for help, but not ready to disappear from the process. They want AI that empowers, not replaces. For the time being, AI shopping will live in a hybrid zone, co-piloting, learning from choices rather than making them outright.
The coming decade will test where that balance lands.
