Agentic Commerce Has a Trust Problem — Here's What Comes Next

Why autonomous AI shopping agents are hitting a wall, and what the industry should be building instead.

The loudest voices in commerce right now are screaming about agentic AI. The pitch is seductive: autonomous agents that browse, compare, negotiate, and buy on your behalf. No more search bars. No more decision fatigue. Just tell the AI what you want and it handles the rest.

There's one problem. The people who actually control commerce — the platforms and retailers — are actively working to make sure it never happens.

The walled garden problem

Eric Seufert, one of the sharpest analysts in digital advertising, has been making this argument forcefully. The fundamental issue with independent agentic commerce is that it directly threatens the business model of every major retail platform.

Amazon and Shopify collectively control over 50% of US e-commerce. Both have been actively blocking external AI agents from accessing their product listings. Amazon updated its site code to ward off AI crawlers from Google and others. Shopify merchants are funneled through Shopify's own ecosystem.

Why? Because discovery is where the money is.

When a customer searches on Amazon, Amazon controls what they see. That control powers a $60 billion advertising business. Every product recommendation, every sponsored listing, every "customers also bought" widget is monetized. Letting a third-party AI agent bypass all of that and go straight to purchase is economic suicide.

The same logic applies to every major retailer. Discovery isn't just a feature — it's the entire business model of modern e-commerce. Retail media networks (the advertising businesses built on top of retail platforms) are the fastest-growing segment of digital advertising. No retailer is going to hand over the keys to an AI agent they don't control.

The Walmart experiment confirms it

The Walmart-ChatGPT saga is the most instructive case study we have so far. In late 2025, Walmart partnered with OpenAI to enable purchases through ChatGPT's Instant Checkout feature. The idea was simple: shoppers could buy Walmart products without ever leaving the chat interface.

It didn't work. Reports indicate the feature struggled with accuracy — wrong items appearing in carts, conversion rates well below what Walmart sees through its own channels. Walmart pulled the plug.

But here's the telling part: Walmart didn't walk away from AI commerce. Instead, they embedded their own AI assistant, Sparky, directly into ChatGPT and Google Gemini. The retailer kept the distribution (hundreds of millions of ChatGPT users) but took back ownership of the shopping experience, the customer data, and the transaction.

The pattern is unmistakable: retailers want AI-powered distribution channels, but they refuse to cede control of the customer relationship. As one analysis put it, Walmart's approach is to "own the agent, rent the distribution."

The real tension nobody is addressing

So we're stuck in a standoff:

The old model is failing. Keyword search doesn't work for how people actually think about products. When someone wants "off-duty model vibes under $80," no search bar on earth returns useful results. 67% of shoppers abandon because they can't find what they want. Over $350 billion is lost annually to poor product discovery.

The new model threatens everyone. Autonomous AI agents that bypass brand relationships, commoditize product catalogs, and disintermediate retailers from their own customers are a non-starter. Platforms won't allow it. Brands won't accept it. And consumers, frankly, aren't ready to hand over their purchasing decisions to an AI.

This isn't a technology problem. It's an incentive alignment problem.

The third way: context handover

What if there's an architecture that gives each party what it actually wants?

Consider a model where the AI owns the messy, subjective, high-effort part of shopping — the initial discovery. Understanding intent. Refining preferences. Filtering options. Learning taste in a 60-second visual conversation instead of a 60-minute browse-and-bounce session.

But instead of trying to own the checkout, the AI packages everything it's learned — the customer's intent, their preferences, their constraints, and crucially, what they've already ruled out — and hands that structured context to the brand.

The brand receives a customer who's already been understood. No cold start. No repeated questions. The customer journey evolves rather than resetting every time they land on a new site.

This is the "context handover" model, and it resolves the tension that's paralyzing agentic commerce:

Brands keep control of the relationship, the data, and the close. They receive a warmer lead than any ad click could deliver, because the AI has already done the qualification work.
The AI earns as a referrer, not a gatekeeper. It's compensated for the discovery value it creates, without trying to own the transaction.
Consumers get better discovery without surrendering purchasing power to an agent they don't fully trust.

Nobody has to give up what matters to them.

Why this matters now

ChatGPT already drives 20% of Walmart's referral traffic. Nearly 60% of US consumers have used generative AI tools for shopping help. The shift in how people discover products is happening — the question is who captures the value.

The agentic commerce maximalists want the AI to capture everything: discovery, comparison, negotiation, purchase, and post-purchase. That's why platforms are building walls.

The context handover model is deliberately restrained. It says: AI is extraordinary at understanding what people want. Brands are extraordinary at fulfilling those wants with personalization, loyalty, and expert guidance. Let each party do what it's best at, with a structured handoff in between.

The companies that figure out this handover — that build the connective tissue between AI-powered discovery and brand-owned conversion — will define the next era of commerce. Not the ones trying to replace the entire funnel with an autonomous agent.

The journey should evolve, not reset. And the brands that embrace this model early will have the warmest customers in the market.

This post references analysis from Eric Seufert (MobileDevMemo), Juozas Kaziukėnas, and industry reporting from Modern Retail, Retail Dive, and Wired.