Shoppers want AI agents to buy products but only under strict rules

The gap between discovery and buying

Consumers are increasingly comfortable using artificial intelligence to find products, but they are much more hesitant to let those tools actually pull the trigger on a transaction. According to a report by Riskified, there is a significant difference between using AI as a research assistant and using it as an autonomous buyer.

The data shows that 80 percent of people polled have already used AI for parts of their shopping journey. However, only 11 percent of those surveyed have actually finished a purchase within an AI tool. This suggests that while the technology is being integrated into the early stages of shopping, the leap to full automation remains a hurdle.

The report notes that consumers expressed openness toward agentic purchases in some form. Despite this general openness, only 7 percent of people are comfortable letting an AI agent buy something without their immediate approval. This pattern indicates that shoppers are embracing AI as a helpful assistant faster than they are accepting it as a substitute for their own decision-making.

Defining the rules of autonomy

For AI agents to become a standard part of commerce, shoppers need to establish clear boundaries on how much authority the software can wield. The research highlights a trend of conditional autonomy, where people are willing to give AI more power only if they can set the specific limits of that power first.

The survey found that 43 percent of respondents still want to approve every single purchase before it happens. Meanwhile, 29 percent said they would be okay with an AI agent making automatic purchases as long as the user sets specific rules, such as receiving notifications after a sale or requiring human approval for exceptions.

This need for governance is a primary barrier to growth in the agentic commerce market. The report's authors suggest that companies must address these open-ended questions of control to clear the path for confident adoption. Without clear terms of use, consumers may remain wary of letting agents handle their money and personal preferences.

Trust and platform responsibility

When it comes to where these agents live, consumers seem to trust general AI platforms more than traditional financial institutions. One-third of those surveyed said they would trust a general platform like ChatGPT or Gemini to run a shopping agent.

In contrast, only 17 percent of people expressed trust in a retailer they already shop with, and just 9 percent would trust a bank or credit card company to manage the agent. This trust is complicated by the fact that security of payment credentials and the AI choosing the wrong product are two of the biggest concerns for shoppers.

The issue of accountability is also a major sticking point for the public. The report found that 51 percent of consumers believe the AI platform should be responsible for issuing refunds if a mistake occurs. This creates a conflict with current terms and conditions from popular agents like Muse and Instinct, which state that the user is responsible for the actions taken by the agent.

Fraud and the merchant's dilemma

The shift toward agentic commerce introduces new risks for the businesses selling the products. Jeff Otto, chief marketing officer at Riskified, noted that while consumers are comfortable delegating shopping, they have not yet delegated the blame for errors.

Because many users expect the AI platform to handle refunds, the dispute often lands on the merchant when a mistake happens. Otto warned that agents could become a new way for people to abuse refund policies. Merchants will need to find ways to distinguish between a real customer using a legitimate agent and a malicious actor using automation to exploit the system.

To solve this, Riskified is expanding its identity risk intelligence to verify the person behind the agent. The goal is to ensure that the person initiating the order is legitimate while still allowing the convenience of automated interactions. This identity-based insight is intended to help merchants reduce chargeback exposure while supporting safer commerce.

Transparency and future growth

As these tools become more common, transparency will be a non-negotiable requirement for the average shopper. The data shows that 68 percent of people want to be told immediately if they are talking to an AI agent rather than a human representative.

This demand for clarity suggests that the industry must prioritize clear communication regarding AI involvement. As agents like Muse and Instinct gain broader attention, the focus will likely shift toward building robust governance frameworks that satisfy these consumer demands for transparency.

Ultimately, the success of agentic commerce depends on balancing convenience with consumer safety. Retailers and brands will need to navigate a landscape where users are willing to delegate tasks, provided they maintain a sense of control and have a clear understanding of who is liable for the results.