AI Wants to Shop For You, But You Still Want Your Credit Card

According to PYMNTS.com, their latest “Agentic AI Report” surveyed 2,299 U.S. consumers from November 10 to December 10, 2025. It found that 71% of people are interested in using autonomous AI for health and wellness management, with 70% interested for travel planning. For commerce, about two-thirds are interested in AI handling groceries, subscriptions, and bill payments. Nearly half of interested consumers would delegate both routine and large, researched purchases to an AI. But here’s the kicker: when these systems actually spend money, consumer trust anchors firmly in traditional finance. Only 3% of traditional, non-AI users trust a generative AI platform as the provider, compared to 26% who trust their bank and 21% who trust a credit card network.

The Trust Chasm

This is the central tension. People are clearly tired of mundane tasks. The report frames this not as novelty-seeking, but as a desperate grab for friction reduction. Who wants to manually track subscriptions or plan weekly meals? So the interest is real. But interest and actual, hands-off trust are two very different things. The data shows that as AI moves from “assistive” (giving advice) to “agentic” (taking action), trust requirements become concrete. It’s no longer about whether the AI gives a good restaurant recommendation. It’s about who’s liable when it accidentally buys 100 pounds of kale instead of one. Consumers are saying they want the AI to have the idea, but they want their bank or credit card company to hold the bag.

Where Does the Sale Actually Happen?

This gets into a fascinating split in consumer behavior. For AI-heavy “Power Users,” the journey is becoming self-contained. 58% of dedicated AI platform users prefer to checkout right inside that AI environment, from discovery to payment. That’s a huge shift. It means for these users, the AI platform is becoming the storefront, the shopping cart, and potentially the payment processor. Merchants will have to become “AI-visible” or risk being invisible. But for everyone else—people using search engines or traditional workflows—the preference is still to finish on the merchant’s own website. Even if AI helped them find the product, they want the final step to happen in a familiar, brand-controlled space. So we’re looking at a bifurcated future of commerce, dictated entirely by which interface a consumer trusts most.

The Path to Scale is Interruptible

Here’s the most crucial design insight from the report: consumers aren’t rejecting autonomy. They’re insisting it be interruptible and accountable. Trust skyrockets when users can approve actions before they happen, reverse decisions, or call for a human override. Basically, they want an undo button and a big red “STOP” sign. The other non-negotiable is data. A significant chunk of users—anywhere from a quarter to nearly half, depending on the segment—will simply block an AI’s access to their personal data. That means product design can’t assume blanket consent. The winning model will be tiered: a “lite mode” with generic help for the skeptical, and a full, data-rich mode for those who opt in. You have to meet users where they are on the trust spectrum.

Winners and The New Middlemen

So who wins in this agentic AI commerce world? The immediate beneficiaries are the entrenched financial players. Digital wallets, banks, and credit card networks are consistently named as the most trusted providers to actually run an AI assistant that spends money. Their regulated, familiar status is a massive asset. The generative AI platforms themselves face a harder climb. They might control discovery and even the shopping interface, but convincing people to let them hold payment credentials is a much taller order. They’ll likely have to partner deeply with those trusted financial rails. The real shift is that the point of competition moves upstream. It’s less about the checkout button and more about which AI agent you delegate your intent to. And for that agent to work, it needs to plug into the old, reliable financial infrastructure we already have. The future is autonomous, but it’s running on very familiar rails.