Shoppers say they'll let AI agents buy for them — but they still want veto power over the final purchase and the brand picked.

This looks at how much say shoppers want over purchases made by AI agents that buy on their behalf: whether they want to sign off on each purchase, set limits ahead of time, or just trust the agent. The short answer from the corpus: people say they're happy to hand over the shopping, but the control they actually want sits in two places. One is the final 'buy' button. The other is the choice of brand.

The headline numbers look like full trust. In Accenture's survey of 25,590 people, 74% said they'd trust an AI agent to make purchases. But the same survey found 56% would still name the brands they want, and 37% of loyal customers wouldn't let an agent switch them to another brand Do consumers actually want to delegate shopping decisions to AI?. RTB House found that the safeguard people want most is approving a purchase before it goes through. 35% asked for it overall, and 44% of baby boomers did Do consumers want to approve AI shopping purchases before they happen?. Put together, people seem glad to let an agent do the searching and comparing, and they want to keep two things for themselves: the choice of brand and the moment money leaves the account, which is hard to undo. Also, what drives people to accept an autonomous shopping agent is how much they expect to gain, not how in control they feel. Acceptance rises when goods are scarce and the agent's speed matters more What drives shoppers to accept autonomous AI assistants?. So the amount of control people demand may shift with the situation rather than stay fixed.

One idea in the corpus makes 'control' harder to pin down. Bilgihan and colleagues argue that an agent which actually completes purchases doesn't just carry out what you already wanted. Through repeated back-and-forth, it helps shape what you want Do AI shopping agents reshape what consumers actually want?. If that's right, approving each purchase checks the final step but not the many small conversations that led there. Naming your brands up front may be people's instinct to guard the part of the decision they feel is still theirs.

Whose control actually counts depends on who built the agent. The 'Coasean Singularity' paper argues that an agent built by a shopping platform inherits the seller's goals by default, while an agent the consumer owns can be set up around the buyer's goals Does agent ownership determine whose interests get optimized?. Bob Hedges warns that agents funded by advertising have a built-in pull toward paid placements, however well they're designed Do advertising-funded shopping agents favor paid product placements?. This creates a tension. Accenture finds consumers prefer agents built by retailers Will brands succeed by building their own AI shopping agents?, which are exactly the agents this research says are least likely to work for the buyer. An approval button only helps if the options it shows you weren't already filtered for someone else's benefit.

That's why part of the corpus argues that controls consumers manage themselves aren't enough. Agency law and economics name three problems with any agent: you can't see what it knows, it uses its own judgment, and its loyalty is uncertain. The usual fixes (incentives, monitoring, enforcement) break down when agents decide quickly, at scale, and in ways nobody can inspect Can traditional governance tools actually control AI agents?. Owens and Law propose holding shopping agents to the same legal duties of loyalty and care that apply to human brokers. They argue this needs to happen before the market's habits harden Should AI shopping agents face fiduciary duties like human brokers?. Their implication: the control consumers want may end up coming more from the rules agents must follow than from a confirm button. One gap remains. Nearly all the evidence here is surveys of what people say they want, not observation of how they actually supervise agents once they use them.