Consumers are increasingly willing to let AI handle shopping tasks, but their trust is selective rather than unconditional. Accenture’s 2026 Consumer Pulse Research surveyed 25,590 people in 16 countries and found strong interest in delegating repetitive work while retaining control over payment and identity-rich choices.
The distinction is important for retailers racing toward agentic commerce. People may welcome software that renegotiates a subscription or reorders groceries without wanting it to choose a holiday, outfit or meaningful gift.
Delegation rises where effort is high and emotion is low
Accenture found that 74% would allow an agent to handle routine activities such as deal negotiation, complaint resolution, renewals and reorders. The same share said they would trust a personal agent more than their best friend to purchase on their behalf.
Trust falls as authority becomes consequential. Thirty-two percent would let an agent make a bounded recommendation for later approval, while 9% would permit a completed purchase without final confirmation. Only 12% were open to autonomous decisions at payment.
Recurring services ranked highly because the desired outcome is usually stable. Lifestyle and travel purchases saw a sharper decline as autonomy increased, reflecting the role of taste, enjoyment and risk.
This suggests a sensible product sequence. Agents should begin with monitoring, comparison and routine service, then earn permission through accurate performance. Full purchasing authority is a separate feature, not the inevitable end of every assistant.
Control and recourse create more trust than intelligence alone
Consumers highlighted safeguards, configurable permissions, immediate overrides, platform reputation, neutrality and clear recourse. Those requirements describe a relationship of delegated authority, not just a better recommendation model.
Users need to specify budget, acceptable brands, timing and when approval is mandatory. They should be able to inspect why an agent chose a product and reverse permission without searching through settings.
Neutrality is difficult when the platform earns advertising or sales revenue. Sponsored products and commissions must be distinguishable from an agent’s independent assessment. Otherwise the assistant may appear to protect the consumer while quietly serving a marketplace.
Data protection is equally central. An agent capable of shopping for an “idealised self” may infer health goals, financial constraints and personal insecurities. Helpful personalisation can become intrusive if those details flow into unrelated advertising or remain indefinitely.
Brands must become legible to both agents and people
Product attributes, stock, price, policies and verified claims need machine-readable structure. Agents may compare a brand before a customer visits its website, changing how discovery and loyalty work.
Fifty-six percent said they would tell an agent which brands to consider, but 37% of behaviourally loyal consumers would allow a switch for a better fit. Loyalty therefore becomes evidence about product, fulfilment and service rather than repeated exposure alone.
Consumers also want agents that cross providers: 61% expressed interest in multi-retailer grocery shopping, and 71% wanted complete trip planning across airlines, hotels and activities. Closed assistants cannot satisfy that expectation unless they cooperate with outside systems.
Physical stores remain important. Most respondents expected AI to change their role, while nearly a third believed stores would gain importance as places of enjoyment. Automation may handle comparison while people seek touch, advice and experience in person.
Retailers should measure complaint resolution, unsuitable recommendations, returns and override frequency alongside conversion. Those outcomes reveal whether delegation creates confidence or merely accelerates transactions.
Agents should also communicate uncertainty. When products are closely matched or important data is missing, presenting alternatives is more trustworthy than manufacturing one definitive “best” choice.
Consumer trust is not a blanket mandate for autonomous shopping. It is permission granted task by task. The agents that succeed will make boundaries obvious, protect personal context and return control before a convenient helper starts to feel like an unaccountable buyer.
