People say they'd trust an AI to shop for them, yet still insist on naming the brand — why the holdout?

This explores whether the effort of explaining your preferences to an AI is what keeps people from handing over purchases completely. The corpus has no study that measures this cost directly. What it does have points somewhere more interesting: the caution may come less from how hard preferences are to state, and more from the fact that preferences often don't exist in finished form until the shopping happens, and from doubt that the agent will stick to what you tell it.

Start with what consumers say. In Accenture's large survey, most people said they would trust an AI agent with purchases. Yet 56% would still name specific brands, and many loyal shoppers wouldn't let the agent switch them to another brand Do consumers actually want to delegate shopping decisions to AI?. If stating preferences were the main burden, you'd expect people to hand off the detail work happily. Instead they keep the most specific choice, the brand, for themselves. The pattern looks less like "explaining is too much work" and more like "I want to keep the decisions that matter to me." A separate line of work finds that acceptance of highly autonomous shopping assistants is driven by perceived benefits, especially under scarcity, rather than by consumers' sense of control What drives shoppers to accept autonomous AI assistants?. That suggests people weigh payoff against risk, not just the effort of setup.

The deeper challenge to the question's premise comes from Bilgihan and colleagues. They argue that with agents that actually complete purchases, preferences aren't fully formed beforehand and then handed over. They take shape through repeated back-and-forth between the agent and the shopper Do AI shopping agents reshape what consumers actually want?. If so, the real difficulty isn't the cost of articulating a preference. It's that a complete specification can't be written down in advance, so full delegation means letting the agent help decide what you want. That is a reasonable thing to be wary of.

The agent-side evidence suggests the wariness is justified. Even when constraints are stated clearly, LLM buyer agents overpaid, revealed budgets they didn't need to, and ignored spending limits across several model families Do LLM negotiation agents reliably follow budget and pricing rules?. In one test, a Gemini agent quietly switched from comparing unit prices to just minimizing the total price as order size grew, and its rate of picking the best option fell from 80% to 35% Does higher purchase volume make AI shoppers search less carefully?. Agents also swing much harder than humans in response to price, rating, and nudge cues Do shopping agents respond more strongly to price and rating cues?. So even a carefully stated preference may not survive execution.

The part you might not expect is that stating preferences well can work against you. A seller-side agent recovered a buyer's real willingness to pay almost perfectly from conversation alone when the buyer's agent was given a descriptive persona. A confidential numeric budget leaked far less Can a seller infer what a buyer really wants to pay?. Related work finds this leakage comes from personalization itself, so better prompting can't fix it Can better prompts stop preference leakage in AI shopping?. Add agents funded by merchant referral fees, whose loyalty is structurally split Can an AI agent serve both merchant and user interests fairly?, and caution about full delegation looks less like reluctance to explain yourself and more like a sensible reading of the risks: the agent may not follow what you said, and what you said may be used against you.