Consumers trust AI to advise, but don’t consent to it purchasing for them

What happened

New research covered by Customer Experience Dive finds a clear split in how shoppers are willing to use artificial intelligence: they are largely comfortable with AI offering product recommendations, but far less willing to let AI actually complete a purchase on their behalf. The findings point to a meaningful trust gap between AI as an advisor and AI as an autonomous agent acting with a shopper's money.

The research frames this as a distinction between "advisory" AI — tools that suggest, compare or narrow down options — and "agentic" AI, which would take the further step of executing a transaction without a human confirming each detail. Consumers appear ready for the former but are holding back on the latter, even as retailers and technology vendors push forward with more autonomous, agent-based commerce tools.

Why it matters

The gap matters because it sits directly on the path many retailers and platforms are currently building toward: agentic checkout, where AI doesn't just guide a purchase decision but initiates and finalises it. If consumer trust hasn't caught up to the technology's capability, businesses risk over-investing in autonomous purchasing features that shoppers aren't yet prepared to hand control to — while under-investing in the advisory layer where trust already exists and value can be captured sooner.

For experience and transformation leaders, the signal is less about whether agentic AI works technically and more about whether customers consent to it. That distinction should shape sequencing: strengthen and scale AI-led recommendation and guidance now, while treating full purchase autonomy as a longer, trust-dependent rollout rather than a simple feature switch.

The Renascence take

This finding confirms something behavioral science has long predicted: trust in a recommendation and trust in a transaction are not the same psychological event, and closing the distance between them can't be engineered purely through better algorithms.

Most organisations will read this as a technology maturity problem — "give it time, trust will follow." It's actually a control and accountability problem. A recommendation carries no consequence if it's wrong; a completed purchase does, and consumers instinctively want a visible moment of human sign-off before money moves, regardless of how accurate the AI is. Operators chasing agentic commerce should design for a visible "last click of consent" as a permanent feature, not a transitional one — because the discomfort here isn't about capability, it's about who bears the risk when AI gets it wrong.