Your Customer Doesn't Want to Talk to You Anymore
Agentic commerce isn't about AI helping customers find you. It's about customers never seeing you at all.
*This article was originally published on rebeccaraebarton.com
There’s a version of the AI-in-commerce story that’s comfortable. Reassuring, even. It goes like this: artificial intelligence will help your customers find you faster. Better recommendations. Smarter search. Personalization at scale. The funnel gets more efficient. Everyone wins.
That’s not the story I want to tell you.
The story I want to tell you is darker, weirder, and probably more true. It’s the story of what happens when your customer stops being your customer at all. When the person making the purchase decision isn’t a person. When the thing you’ve spent years building — brand awareness, emotional connection, carefully cultivated loyalty — becomes irrelevant because an algorithm is doing the shopping.
Welcome to agentic commerce. The era when AI doesn’t just help people buy things. It buys things for them.
Earlier this month (Jan 2026), Google announced something called the Universal Commerce Protocol at NRF. The name is deliberately boring, the kind of thing that slides past you in a press release. Don’t let it. UCP is an open standard that lets AI agents — not humans, agents — handle the entire shopping journey. Discovery. Comparison. Checkout. Payment. The protocol was co-developed with Shopify, Target, Walmart, Wayfair, and Etsy. It’s endorsed by Visa, Mastercard, American Express, Stripe, and Adyen. This isn’t a pilot. This is infrastructure.
OpenAI already has checkout inside ChatGPT. You can buy things without leaving the conversation. Perplexity has been doing the same with PayPal. Amazon launched “Buy for Me” — an agent that purchases products from other websites on your behalf. The company that built its empire on being the everything store is now building tools to buy from everywhere else, because they understand where this is going even if most brands don’t.
Here’s where it gets uncomfortable: about a third of U.S. consumers say they’d let an AI make purchases for them. Not help them decide. Make the purchase. Autonomously. And that number is growing fast among younger cohorts who already trust recommendation algorithms more than they trust advertising.
Morgan Stanley projects that by 2030, nearly half of online shoppers will use AI shopping agents, accounting for roughly 25% of their spending. Traffic from AI sources to retail sites is up 1,200% year over year, according to Adobe. The traditional funnel — awareness, consideration, conversion — doesn’t just get shorter. It disappears. The customer never sees your homepage. Never browses your collection. Never reads your about page or watches your founder video or feels the subtle emotional pull of your brand story.
They ask their agent for “running shoes under $150 that work for flat feet.” The agent returns three options. The customer picks one. Or doesn’t even pick — just approves the agent’s recommendation. Transaction complete.
What does “brand” even mean in that interaction?
I’ve spent a lot of my career thinking about brand as differentiation. As meaning. As the accumulated weight of every touchpoint, every experience, every moment where you made someone feel something. Brand as moat. Brand as the reason someone picks you over the cheaper alternative.
But AI agents don’t feel. They optimize. They process attributes: price, reviews, specifications, availability, return policies. They might factor in “brand sentiment” as a data point, but it’s just that — a point. One variable among hundreds. The emotional resonance you’ve spent years building becomes a line item in a feature comparison.
Amazon understands this, which is why they’ve blocked outside AI agents from accessing their site. They’re protecting their $56 billion advertising business — all those sponsored placements that depend on humans browsing, seeing, clicking. If an AI agent is doing the shopping, nobody sees the ads. The entire economic model of product discovery collapses.
They even sued Perplexity for scraping their site. That lawsuit isn’t really about data access. It’s about who controls the customer relationship in a world where customers outsource their decisions to machines.
The optimistic take is that this forces brands to focus on what actually matters: product quality, genuine differentiation, the kind of objective excellence that survives algorithmic scrutiny. If an AI agent is comparing your product to ten alternatives on pure merits, you’d better have merits. No amount of clever positioning or emotional branding will save a mediocre product when a machine is doing the evaluation.
The pessimistic take is that most brands aren’t ready for that. They’ve been relying on the fuzzy middle — good enough product, strong enough marketing, sufficient brand recognition to win the consideration set. That fuzzy middle gets obliterated when an AI agent can evaluate every option in a category in seconds and surface the objectively best choice for a given set of criteria.
And then there’s the question I’m not seeing asked often enough: how will these agents be influenced?
On one side, this could force a genuine meritocracy. If an AI agent is evaluating your product on structured data — specs, reviews, return rates, price, availability — you can’t charm your way through a mediocre offering. The brands that invest in product quality, clean data feeds, transparent pricing, and honest reviews will be the ones the agents recommend. That’s not a bad outcome. It rewards the things that should have mattered all along. On the other side, every system designed to reward merit eventually gets gamed. Brands are already stuffing structured data feeds with inflated specs and keyword-dense descriptions because agents parse attributes, not brand stories. Review manipulation — already a billion-dollar problem — becomes existential when the difference between a 4.6 and a 4.4 is the difference between being recommended and being invisible. Dynamic pricing algorithms are being tuned to win agent comparisons in milliseconds, not to persuade a human browsing a page. Security researchers have already demonstrated that product listings can carry hidden instructions that hijack what a shopping agent sees, ignores, or recommends — prompt injection is ranked the number-one vulnerability in AI applications by OWASP, and it’s tailor-made for commerce. And the platforms themselves have every incentive to sell placement inside the agent’s recommendation set.
Google’s UCP, Amazon’s Buy for Me, ChatGPT’s Instant Checkout — the moment one of them starts charging for priority ranking, we’ve rebuilt the pay-to-play ad economy, except the customer never even sees the ad. They just get the output. This is SEO with AEO sprinkles. The first generation of search rewarded the best content — until it didn’t. Until keyword stuffing, link farms, and paid placement turned it into a pay-to-play system that Google has spent two decades trying to unwind. Agentic commerce is starting that same cycle, except the stakes are higher. With Google search, the customer still made the final decision. With agentic commerce, the agent decides. The customer approves. Maybe.
The protocols are live. The agents are shopping. The rules are being written by the companies that stand to profit from them. If your brand’s strategy still assumes a human on the other end of the funnel, you’re building for a world that’s already disappearing.
Your customer doesn’t want to talk to you (or your agent) anymore. They want to talk to their agent. And their agent wants to talk to your data feed.
The question is whether you have anything worth saying.
