While retailers are sprinting toward autonomous AI. Shoppers are asking for something much smaller: a second opinion.
Two reports published in the past few weeks make that gap hard to ignore. The Harris Poll's Algorithmic Aisle study (June 2026, surveying 3,222 consumers across the US, UK, Brazil and India) found that shoppers are genuinely comfortable with AI in their shopping journey, comparing prices, summarising reviews, and flagging dubious claims.
What they are not comfortable with is AI making the call for them.
Meanwhile, Salesforce's fourth-edition State of Commerce report (surveying 3,450 commerce professionals across 20 countries) shows the industry racing in the opposite direction: toward agentic AI, automation at scale, and AI-run decisioning, often faster than the data and strategy underneath it can support.
Read side by side, the two reports describe the same moment from opposite ends of the counter. One is about what people want. The other is about what businesses are building.
The distance between them is where the next 18 months of retail AI will actually be decided.
What shoppers will actually let AI do
The framing is precise: 72% of consumers globally are comfortable with AI helping them shop, provided they still make the final decision.
That comfort is highest for exactly the tasks you'd expect a helpful assistant to handle: comparing prices across brands or retailers (74%), checking whether a claim is credible, summarising reviews, and suggesting products based on past purchases (71% each).
Comfort drops sharply the moment AI stops informing and starts deciding.
Only 64% trust AI to recommend when they should switch brands.
Just 58% are comfortable with AI automatically reordering products they buy regularly.
The pattern holds across every category in the study: shoppers want AI to narrow the shelf, not restock the pantry without asking.
Trust, where it exists, is conditional and has to be earned. 72% of consumers worry AI will narrow what they see rather than expand it. 74% say they'd grow sceptical if an AI tool kept steering them toward premium or expensive products. And 78% assume, correctly or not, that brands are paying their way into AI-generated recommendations.
The fix, according to the same shoppers, isn't complicated. 31% say a recommendation feels more trustworthy when it shows both pros and cons. 29% want a plain explanation of why a product was suggested. Verified reviews help too.
What retailers are actually building
Set that against the Salesforce data and the mismatch comes into focus.
78% of commerce organisations already use some form of AI. Fewer than a third currently use agentic AI, but nearly half plan to deploy it within six months.
Many ecommerce professionals agree that scaling across channels isn't viable without AI at all. The direction of travel is unmistakably toward more autonomy.
The pressure driving that shift is real. 86% of commerce professionals say AI is raising the bar for customer expectations while 61% say meeting those expectations is harder than ever.
Teams are essentially being asked to do more with less. AI is simultaneously the source of the pressure and the only credible answer to it, which is exactly why organisations are moving fast.
But it’s clear what's underneath that speed. Poor data integration, an undefined AI strategy, or poor data quality are the major barriers many organizations face.
The study reports only 27% say their customer data is fully unified across sales, service, marketing and commerce while only 32% have fully defined AI success metrics.
Organisations are scaling AI before they've agreed on what winning looks like, and before the data feeding it is reliable enough to earn the trust shoppers are already withholding.
The discovery layer is moving just as fast. Organisations are reporting a rise in traffic from LLM-powered search, with consumer reliance on AI assistants as the first stop in the shopping journey up 200% in a single year (May 2025 to May 2026).
Retailers are adopting AI faster than their foundations can support
The gap is the opportunity
Put the two reports together and the shape of the problem is obvious.
Shoppers have told researchers, in detail, what they'll accept from AI: comparison, explanation, summarisation, and recommendation with visible reasoning.
What they've explicitly said no to is silent automation, unexplained bias, and undisclosed sponsorship.
Commerce leaders, meanwhile, are moving toward exactly the kind of automation shoppers are wary of, often without the unified data or defined success metrics that would let them do it credibly.
This isn't an argument against agentic AI or automation. It's an argument for sequencing.
The organisations that will win the next phase of retail AI aren't the ones deploying the most autonomous agent the fastest. They're the ones building AI that shoppers can see through: grounded in real product data, transparent about why it's recommending something, and clear about where a human takes over.
That's the model we've built Clevertar's AI Sales & Support Agents around from the start, not because we anticipated this exact research, but because it's what actually works in a live retail deployment.
Every recommendation is grounded in the client's real product catalogue and policies, not free-generated. And because the agent is trained and scoped to a single brand's voice and inventory, there's no ambiguity on where the information is coming from or if it’s trustworthy.
Sources
The Harris Poll, "What Do Shoppers Really Want From AI?" (The Algorithmic Aisle, June 2026)
Salesforce, "State of Commerce Report," 4th edition (survey fielded April 10 - June 4, 2026)
