Consumers trust AI advice but resist letting it buy for them

What happened

New reporting from Customer Experience Dive finds a split in how consumers feel about artificial intelligence in their shopping journeys: many are comfortable turning to AI tools for product recommendations and advice, but far fewer are willing to let those same tools complete a purchase on their behalf. The gap points to a trust boundary that sits between AI as an assistant and AI as an agent with spending authority.

The distinction matters as more retailers, banks and service providers experiment with AI-driven "agentic" shopping features — from conversational recommendation engines to autonomous checkout bots. According to the reporting, consumers are broadly receptive to AI helping them decide what to buy, but grow markedly more cautious once AI is positioned to execute a transaction using their money or payment credentials.

Why it matters

For brands racing to deploy agentic AI in commerce, the finding reframes the adoption question. The barrier to scaling AI-led purchasing isn't primarily technical capability — it's consumer consent and control. Organisations building AI shopping assistants need to design for a clear handoff point where the customer, not the algorithm, makes the final call on spending.

This has direct implications for experience design, trust architecture and regulatory posture. Companies that blur the line between advisory AI and transactional AI risk a backlash if customers feel money was spent without sufficiently explicit permission — even if the AI's recommendation was accurate. Getting the consent moment right may matter more to adoption curves than getting the AI's judgement right.

The Renascence take

This is a textbook case of the difference between cognitive trust and behavioral trust. Customers can intellectually trust an AI's recommendation while still refusing to hand over agency — because the two trigger completely different psychological risk calculations: being wrong versus losing control.

The mistake most operators will make is treating this as a temporary trust gap to be closed with better AI accuracy or more reassuring UX copy. It isn't. Consent over spending is a control preference, not a confidence problem, and no amount of AI performance will fully dissolve it. The smarter move is to design the "last click" as a deliberate, visible ritual — a moment where the customer explicitly authorises the AI's suggestion — rather than trying to engineer that friction away. Brands that rush to full autonomy in the name of convenience are optimising for the wrong metric; the ones that win will treat the human authorisation step as a feature, not a bug.