Consumer AI Adoption Moves Faster in Shopping Than at Checkout

The shopper is changing before the checkout does

The usual story about AI and supply chains begins in the warehouse. It covers forecasting engines, rerouting software and autonomous agents that rebalance stock. This year's coverage points to a different starting place: the consumer. People are adopting AI shopping tools quickly for research and slowly for spending money. That gap between discovery and delegation is likely to shape the next phase of supply chain change more than any single back-office tool.

The argument here is that the near-term disruption to supply chains is a change in how demand forms, not a takeover by robot buyers. AI-assisted shoppers compare more options, chase lower prices, try unfamiliar brands and build larger baskets. Each of those habits makes demand less predictable in ways older planning systems were never built to handle.

How many consumers use AI to shop

Every survey agrees that consumer use of AI for shopping is now mainstream. The surveys disagree on exact figures. NielsenIQ's US tracker found that 42% of consumers had used at least one AI tool to shop in the past month.24 Locus surveyed more than 1,000 US online shoppers in the second quarter and put the share at 45%.26 A Commerce and PayPal study found that only 28% of Americans use AI tools specifically when shopping, even though 66% use AI in daily life.28 A multi-country survey by RTB House reported that 68% of respondents had used at least one AI platform to shop in the past three months.30

Europe looks similar. Deloitte surveyed 13,500 European consumers and found that more than half already use AI for shopping. Use is uneven by country: Spain, Poland, Hungary and Italy are above 60%, while the UK, France and Germany sit roughly 20 points lower.21 McKinsey's 2026 European survey found that 38% of consumers use AI to research products or guide purchases.23

The spread of these numbers comes from different definitions and samples. The direction is the same in all of them. Traffic data, which does not depend on what people say in surveys, supports that reading. Adobe Digital Insights recorded a 393% year-over-year jump in AI-referred retail traffic in the first quarter of 2026. Visitors from AI sources during Prime Day converted 40% better than visitors from other channels.23

Where adoption stops

The surveys agree even more closely on where adoption stops. Only 5% of US consumers in NielsenIQ's data have let a fully autonomous agent place an order for them.24 Gartner found that willingness to let AI make purchase decisions peaked at 11%, and that was in low-stakes categories like household supplies and personal care.27 In Deloitte's European data, 24% to 34% of shoppers are willing to hand off comparison and price-finding, but only 8% would hand off checkout.21

Worldpay's seven-market study shows the same pattern from another angle. Among consumers open to AI shopping, 80% would let an agent play some part in payment. But 59% want to approve every purchase themselves rather than give an agent standing authority.22 Visa's CEO said in September that about three-quarters of US consumers do not trust agentic platforms to pay on their own. That figure reverses when a trusted payment brand is involved.21

Safeguards change behavior. RTB House found that 42% of US millennials would let an agent buy for them within a $250 budget if they could return the item within seven days. Without that guarantee, the share drops to about a third.30 Forrester's mid-2026 view is blunt: most so-called agentic experiences are still conversations, and the market story is running well ahead of what shoppers actually do.21

There are also signs of trouble with accuracy. Among Gartner respondents who used AI on a recent purchase, 54% said they had to double-check everything the tools told them, and 62% said the information wasted their time.27 That does not stop adoption, but it does limit how much people will hand over.

Why consumer behavior reaches the warehouse

This is where the supply chain impact begins, and it is mostly indirect. AI-assisted shoppers act differently in ways that weaken the assumptions behind forecasting.

First, brand loyalty becomes less stable. Locus found that 39% of AI-assisted shoppers were more likely to try brands they would not otherwise have considered.26 In RTB House's survey, 59% of US consumers said AI platforms had shown them brands they did not know.30 For planners, that means demand spreads across a wider set of suppliers, and the historical sell-through data that forecasts depend on becomes less reliable.

Second, price matters more. PYMNTS Intelligence found that 46% of AI-assisted shoppers used the technology mainly to find the best price. Another 53% said a lower price could pull them away from a large marketplace to a brand's own site.25 When an agent compares prices constantly, small price changes can move volume quickly, and inventory has to respond quickly as well.

Third, order sizes change. Locus reported that 37% of AI users buy more items per order. The firm notes that this affects fulfillment and how shipping costs are spread across items.26

Merchants already seem to expect a split market. Only 28% of merchants surveyed by PYMNTS would give AI agents their full range on the same terms as other channels. Others would offer agents different prices or delivery options, or only selected products.25 If that happens widely, retailers will need to manage a separate agent channel with its own pricing and availability rules. That is a planning problem as much as a marketing one.

Business adoption is ahead of consumer delegation

Companies are moving faster toward agents than consumers are. The Hackett Group found that 83% of organizations have deployed or are piloting AI in supply chain analytics.8 Gartner expects spending on supply chain software with agentic AI to grow from under $2 billion in 2025 to $53 billion by 2030. It also expects enterprise adoption of agentic features to rise from 5% today to 60%.1 IDC predicts that 45% of the world's 2,000 largest companies will use agentic supply chain orchestration by 2029.5

Some deployments are already running. Lenovo has connected order-fulfillment and risk-management agents directly to its transaction systems across 180 markets. A Fujitsu and Rohto Pharmaceutical exercise cut transport costs by up to 30% in a virtual network, and a live trial runs through March 2027.3 One industry blog credits Amazon's multi-agent system with a 30% increase in same-day deliveries in 2025.11 That figure comes from a vendor-oriented source, not an audited disclosure.

Retail is likely to be the main testing ground. One analyst quoted in NVIDIA's retail survey argued that agentic AI will disrupt supply chains and operations before it disrupts the storefront, because that is where returns can be measured.12 Bain research found that shoppers trust retailers' own on-site agents three times more than outside agents.16 That gives retailers an incentive to keep the customer relationship and the demand signals that come with it.

Spending is not the same as results

The coverage also shows a large gap between adoption and results. Schneider Electric notes that 73% of logistics companies now run at least one AI workflow. Citing BCG, it adds that only 10% of providers report measurable financial impact.2 PwC warns that most digital efforts will fall short by 2028 because companies underinvest in people and processes. Its view is that AI will improve visibility faster than it improves decisions until data and decision rights are fixed.1 The Hackett Group calls the gap between ambition and execution a competitive risk in itself.8

The sources that describe live agent systems all describe tight limits. Agents reroute freight only within cost caps. Large inventory changes pause for a human to approve them. Agents that communicate with suppliers stay in draft mode until they prove accurate.3 This closely matches the consumer side, where shoppers want final approval before a purchase.

What comes next

The most likely near-term outcome is not autonomous shopping at scale. It is a period in which AI shapes what consumers consider and buy, while humans on both sides keep approval at the point of commitment. Projections such as McKinsey and ICSC's estimate of a $1 trillion US agentic commerce market by 2030 assume much more delegation than consumers show today.30

Liability is still unsettled. In the PYMNTS survey, 93% of merchants said the agent provider should cover losses when an agent makes a wrong choice.25 Until that question is resolved, autonomous checkout will stay limited.

The companies that benefit will be the ones whose planning systems can absorb AI-driven demand that is more sensitive to price and less loyal to brands, without waiting for consumers to fully trust agents with their money. Consumer AI is already changing what flows through supply chains, even though it is not yet completing many purchases.