Do shoppers trust AI agents to buy for them? Mostly not yet. In Accenture’s 2026 Consumer Pulse, 74% of consumers would delegate routine tasks to an agent, 32% would let one decide what to buy, and only 9% are open to an agent making fully autonomous purchases (Accenture, “Talk to my AI agent”, June 3, 2026).

For a merchant, that gap shapes where agent readiness pays off first: in being chosen, not in being charged automatically.

What Accenture measured

The survey covered 25,590 people in 16 countries. Accenture describes three levels of delegation, and the share of consumers drops at each one:

  • Task execution (74%). The agent negotiates a deal, resolves a complaint, renews a subscription or reorders, acting strictly on the consumer’s instructions.
  • Delegated decisions (32%). The agent picks the best option within a budget, preferences and an occasion. The consumer reviews and approves, and makes the payment themselves.
  • Autonomous purchasing (9%). The agent initiates and completes purchases on its own within defined rules.

Accenture adds that payments remain the biggest barrier, and that about one in three consumers say a successful low-risk purchase would raise their comfort. It frames the 9% as something that “could grow fast”. That is Accenture’s reading, not a measured trend.

What this means for a store

The numbers describe stated openness in a survey, not observed behaviour. Read that way, they point to three practical conclusions:

  • The agent shortlists first. Accenture’s report says that when agents build the shortlist, “the moment of choice moves upstream to the decision layer”: the standards, data and verification logic that determine what gets chosen. A store that is not readable never reaches the shortlist, whoever pays.
  • Agents reward verifiable facts. The report lists “verified claims, clean product data, transparent pricing, reliable availability and successful fulfillment” as what agents trust. Those are fields on your product pages, not brand copy.
  • Reorders are the first realistic use. The report’s own examples of task-level delegation include reordering and renewals. Stores selling repeat-purchase goods have the most to gain early. This is our reading of the report, not a claim it makes about categories.

Human handoff still matters in the meantime. Adobe’s data on AI-referred visits, covered in our earlier piece on AI referral traffic and conversion, shows shoppers arriving from assistants and finishing the purchase themselves.

What to check on your own store

  1. Can an agent read price, availability and variants from the product page without running JavaScript?
  2. Does the purchase path hold up when a non-human client walks it? Our guide to why AI agents abandon checkout lists the usual breakpoints.
  3. Are returns, shipping cost and delivery time stated in machine-readable form, so an agent can verify your promise?

FAQ

Do consumers trust AI agents to make purchases?

Only a small share do today. In Accenture’s 2026 survey of 25,590 consumers across 16 countries, 9% are open to an AI agent making fully autonomous purchases, against 32% who would let an agent decide what to buy while they pay, and 74% who would delegate routine tasks.

Does this mean agent readiness can wait?

No. Even where the shopper approves and pays, the agent decides which stores make the shortlist. Accenture says the moment of choice moves to the decision layer, which rewards data an agent can verify. Readiness decides whether you are considered, regardless of who completes the payment.

Is the 9% a forecast?

No. It is the share of surveyed consumers who said they are open to autonomous purchasing, as reported by Accenture on June 3, 2026. It measures stated willingness, not purchases made, and Accenture notes that a successful low-risk purchase could raise comfort for about one in three consumers.

Sources

The transaction, not the click, is where a store proves it is ready for agents: shoppers may keep the payment for now, but agents already decide who gets the sale.