AI trust gap: shoppers accept advice, resist autonomous purchases

What happened

New consumer research reported by Customer Experience Dive finds a clear split in how shoppers feel about artificial intelligence in retail: they are largely comfortable having AI recommend products, but far more reluctant to let AI actually complete a purchase on their behalf. The findings point to a meaningful trust gap between AI acting as an adviser and AI acting as an autonomous agent with authority to spend money.

The research frames this as a distinction between "advisory" AI — tools that suggest, filter or personalise options for a human to choose from — and "agentic" AI, which takes the further step of executing a transaction. While retailers and technology vendors have been racing to build agentic shopping assistants, the study suggests consumer appetite has not caught up with the technical capability, particularly at the final, money-changing-hands stage of the journey.

Why it matters

For experience and digital transformation leaders, the story is a useful corrective to the assumption that comfort with AI recommendations automatically translates into comfort with AI autonomy. Retailers investing in agentic commerce — AI that can browse, select and checkout without a human confirming each step — may be building ahead of where trust currently sits. That has direct implications for adoption curves, conversion rates and how much control needs to stay visibly in the customer's hands.

The gap also reframes the design challenge: it is not simply "how good is the AI at predicting what I want," but "how much authority am I willing to hand over, and at what point in the journey." That is a behavioural and trust question as much as a technical one, and it means the rollout of agentic AI in commerce will likely need to be sequenced carefully, with clear opt-ins, visible checkpoints and easy overrides, rather than assumed as a natural next step from recommendation engines.

The Renascence take

The headline risk here isn't that AI recommendations are working — it's that many organisations are treating "customers accept AI advice" as license to push straight to "customers accept AI transactions," when those are psychologically different asks with different risk profiles for the person involved.

Recommendation and payment sit on opposite sides of a trust threshold: one shapes a choice, the other removes it. Customers will delegate judgement long before they delegate control over their money, because the cost of a bad suggestion is a wasted click, while the cost of a bad autonomous purchase is a wasted transaction they didn't approve. Operators building agentic commerce should treat checkout as a deliberate, visible consent point — not a feature to automate away — and earn the right to remove that step only after customers have repeatedly trusted the AI's advisory judgement first.