Shoppers Trust AI Agents With Groceries First — and Little Else
Three new surveys agree: 28% would let an AI buy groceries, 6% luxury goods, and half would not delegate anything. The trust ladder is steep — and it tells merchants exactly where agentic commerce starts.
Three separate surveys tried to measure the same thing this month: how much are consumers actually willing to let an AI agent buy on their behalf? The answers, compiled by Supermarket News, line up into one of the clearest pictures of agentic commerce we've seen — and it's a much narrower mandate than the hype suggests.
The numbers
Nuvei (payments infrastructure) found a steep trust ladder:
| Category | Would let AI buy it |
|---|---|
| Groceries | 28% |
| Household essentials | 21% |
| Subscription renewals | 19% |
| Travel bookings | 9% |
| Luxury goods / financial products | 6% |
| Nothing at all | 51% |
Invoice Home's Agentic Commerce Perception report found the same shape with different ordering: 49% most want to offload household essentials, 39% groceries, 37% toiletries — and 30% said they'd trust Amazon's Alexa most to do the buying.
Koddi found the most optimistic slice: 55% would welcome AI automatically reordering everyday items, and 64% are open to AI recommending brands they wouldn't have considered.
Reading the ladder honestly
The pattern across all three studies is unambiguous: willingness to delegate is inversely proportional to how much the purchase matters. Milk and paper towels, yes. Flights, rarely. Money and jewelry, almost never. And a majority — 51% in Nuvei's sample — won't hand over anything yet.
One reading, popular in commentary on these studies, is cynical: this isn't trust, it's indifference — people delegate what they've stopped caring about. There's truth in that, but it misses how delegation has always worked. Nobody trusted the dishwasher with the crystal on day one either. Low-stakes, high-frequency, easily-reversed purchases are the standard on-ramp for every automation technology — the correct question isn't why the numbers are low, it's how fast the ladder gets climbed. Koddi's 64% openness to new-brand recommendations suggests the answer is: faster than the 6% luxury figure implies. Consumers aren't rejecting agents; they're staging them.
What merchants should actually do with this
- If you sell repeat-purchase goods, agentic readiness is now near-term work, not futurism. A quarter to half of consumers are ready to delegate your category first. That means structured product data, stable SKUs, machine-readable pricing and clean APIs — the things an agent needs to buy from you reliably. (We cover the checklist in how AI shopping agents interact with e-commerce.)
- If you sell considered purchases, the agent's role is research, not checkout. With 6–9% willing to delegate travel and luxury buying, the near-term battle is being recommendable — appearing accurately when a shopper's agent assembles options a human will approve. That 64% openness to unfamiliar brands cuts both ways: agents can introduce you to shoppers who'd never have found you, or route around you entirely.
- The 51% is the market. Half of consumers letting agents buy nothing isn't a rejection — it's the pool every trust-building mechanism (spending caps, approval steps, easy reversals) is competing to convert. The platforms that make delegation feel revocable will convert it first, which is why agent guardrails are a commerce feature now, not just a safety one.
Trust in shopping agents is being built from the bottom of the receipt up. The brands that treat the grocery basket as the proving ground — rather than waiting for consumers to hand over the whole wallet at once — are the ones that will be standing on the upper rungs when the climb happens.
