Two weeks ago, the story here was a gap between what AI could already do in discovery and what shoppers would still let it do at checkout. This week didn't close that gap so much as raise the stakes on both sides of it. NielsenIQ's monthly Agentic Commerce Tracker found that 51% of U.S. consumers have now used at least one AI-powered tool to help them shop in the past month — the first time that figure has crossed the halfway point. That's the adoption half of the story. The other half arrived within days: regulators, banks, and the platforms themselves all moved, separately, to make sure a majority of AI-assisted shopping doesn't mean a majority of unaccountable shopping.
Adoption Just Became the Default, Not the Exception
NIQ's tracker breaks the 51% down by function rather than treating it as one behavior: 20% of shoppers now use AI-powered product recommendations and 16% use a dedicated AI personal shopping assistant, spanning the full path from discovery through purchase decision. NIQ's North America president, Liz Buchanan, called it "a defining moment for the industry," and the framing is fair — this is no longer a story about early adopters. It's a story about the median shopper.
That milestone landed in the same week Adobe forecast U.S. holiday online spend will hit a record $275.1 billion, up 6.7% year over year — the busiest shopping season on record arriving just as AI crosses into majority use. Volume and adoption are compounding at the same time, which is exactly the condition under which accountability questions stop being theoretical.
The FTC Closed the Door on "The Agent Did It"
Those questions got their clearest answer yet from Washington. Speaking at a Reuters event, FTC Chairman Andrew Ferguson rejected the idea that an AI agent is an independent actor that can absorb blame on a company's behalf. His standard, as he described it: regulators will look at "the instructions people give these tools and the companies behind them" — not the tool itself. Ferguson noted that in incidents initially blamed on autonomous AI behavior, closer review typically found the system had simply followed the instructions and permissions it was given. He said he'd rather apply existing law than write AI-specific rules, which in practice means a merchant's documentation of what an agent was authorized to do — and what it wasn't — becomes the evidence that decides liability.
The instructions people give these tools and the companies behind them.
— FTC Chairman Andrew Ferguson, Reuters event, Sept. 25, 2026
Six Major Banks Want a Paper Trail for Every Agentic Purchase
The financial industry reached a similar conclusion from the opposite direction. ASB Bank, Bank of America, Capital One, Commonwealth Bank of Australia, ING, and NatWest jointly published a paper, "Building Trust in Agentic Commerce," calling for every AI-assisted purchase to carry a full audit trail — the customer's original instruction, what authority they granted, the agent's decision, and the payment outcome. The banks' stated problem is a visibility gap: issuers and acquirers currently can't see who an agent is, what merchant it transacted with, or what the shopper actually intended, which leaves everyone stuck when something goes wrong — an overspend, a wrong item, a disputed charge — and no one can say who authorized what. As the paper puts it, "trustworthy agentic commerce will require participants to reconstruct the path from a customer's instruction to the resulting transaction, and to use that record when something goes wrong." The framework is voluntary for now, with a follow-up paper promised to turn the principles into an actual protocol.
Amazon Drew Its Own Line — And a Competitor's Agent Was on the Wrong Side of It
Platforms aren't waiting for regulators or banks to set the terms either. Amazon blocked Meta's Muse AI agent from its marketplace, saying Muse had been accessing Amazon without identifying itself and, more seriously, had been capturing and storing shoppers' Amazon credentials. Amazon's position was that third-party agents "should operate openly and respect service provider decisions," comparing the standard to how delivery and travel-booking apps already work with the platforms they run on top of. It's the latest move in a pattern — Amazon has spent the past year suing Perplexity over its Comet agent and pushing back on agents from Google and OpenAI — and notable here because Meta is simultaneously a major advertiser on Amazon's own marketplace and a customer of its cloud infrastructure. Even that relationship didn't buy Muse an exception.
What This Means for CPG Brands
Three things follow from a week where adoption and accountability moved together. First, "the agent did it" is now explicitly not a defense — per Ferguson's standard, any brand using AI agents in marketing, customer service, or commerce needs to be able to show exactly what that agent was instructed and permitted to do, before it needs to. Second, the audit-trail push from banks is a preview of what merchants will eventually be asked to provide on their own systems — instruction, authorization, decision, outcome — and brands whose product and pricing data isn't structured well enough to produce that trail today are the ones who'll scramble when it's required tomorrow. Third, platform gatekeeping is intensifying, not settling — Amazon's move against Muse shows that which agents even get to represent your brand to a shopper is still being decided by the platforms, not by brands or consumers, which makes staying informed about where and how your products actually appear inside agent conversations a live, weekly concern rather than a one-time audit.
The Bottom Line
A month ago the question was whether shoppers trusted agents enough to use them. NielsenIQ's numbers say that question is largely settled — a majority already do. This week reframed the harder question underneath it: who's responsible when an agent gets it wrong. The FTC says it's the company that gave the instructions. The banks say it's whoever can't produce the record. Amazon says it's whoever doesn't play by the platform's rules. None of those answers is final, but all three arrived in the same week adoption crossed 50% — which means the accountability era of agentic commerce didn't wait for the industry to be ready for it.
Sources: NielsenIQ, "Majority of U.S. Consumers Now Use AI to Shop, NIQ Finds," Sept. 24, 2026; Adobe Newsroom, "U.S. Holiday Shopping Season to Hit Record $275.1 Billion Online," Sept. 28, 2026; PYMNTS.com, "FTC Chair Says Companies Cannot Blame AI Agents for Their Actions," Sept. 28, 2026; PYMNTS.com, "Banks Call for Audit Trail From AI Shopping Instructions to Payment Outcome," Sept. 22, 2026; PYMNTS.com, "Amazon Bars Meta's Muse AI Agent From eCommerce Marketplace," Sept. 21, 2026.