The Three Structural Reasons Retailers Are Resisting AI

The AI commerce narrative of 2024 and 2025 had a broadly optimistic arc: capabilities arriving, infrastructure building, some early data landing to validate the investment thesis. There were sceptics, and some of those sceptics were right about specific things, but the general direction of travel felt like "more, faster."

The 2026 version of the conversation is more complicated. eMarketer's characterisation of retailer pushback as a potential cloud over 2026's progress is accurate, and it's worth being precise about what the pushback actually is. "Backlash" can mean anything from legitimate structural concern to implementation fatigue to plain conservatism, and the three are not equivalent.

Some of what's driving the pushback is structural concern of the most serious kind.

The Margin Erosion Problem

Agentic commerce poses a direct challenge to retail media revenue. AI shopping agents that optimise for lowest price and fastest delivery (which is what a lot of consumer-delegated purchasing ends up doing) systematically bypass sponsored product placements, display advertising, and the paid discovery layer that retailers have built significant revenue structures around.

If you're Tesco or Argos and you've built a retail media network on the back of your ecommerce traffic, and a meaningful share of that traffic starts arriving via AI agents that have already decided what to buy before landing on your site, the advertising revenue attached to the discovery journey evaporates. The transaction still happens, but you've lost the margin contribution from the media layer.

This is a real economic concern, not a theoretical one. Amazon, whose retail media business contributes to the tens of billions US advertisers spend on retail media annually, faces the same pressure. Its AI shopping assistant (Rufus when this article was first published; rebranded as Alexa for Shopping in May 2026) is simultaneously a bet on capturing AI commerce volume and a potential cannibalisation of paid search revenue within Amazon's own ecosystem. The same dynamic plays out in the UK, where Tesco's Media and Insight Platform, Sainsbury's Nectar360 Pollen, and Boots Media Group have all built profitable advertising networks on the back of their first-party purchase data. All three face exactly the same agentic commerce pressure on the search-to-purchase pathway.

The retailers who've built the most profitable retail media networks have the most to lose from a world where AI agents skip the discovery layer. Reconciling that with the agentic commerce opportunity is a genuine strategic tension, not a failure of imagination.

The Customer Relationship Question

There's a version of agentic commerce where the customer's primary AI relationship is with your brand. That's the Walmart approach: embedding your own AI into the commerce experience rather than ceding that interface to a third party. And there's a version where the primary relationship is with a third-party platform (ChatGPT, Perplexity, Google AI Mode) and your products happen to appear in it.

The concern among retailers isn't primarily about whether they make the sale. It's about whether they maintain the customer relationship that enables retention, cross-sell, loyalty programme participation, and the first-party data collection that feeds personalisation and planning. If the customer's relationship is with OpenAI and they happen to buy your product through it, what do you know about why they chose you? What conversation context do you have access to? What's your ability to build on that interaction?

The Universal Commerce Protocol (UCP) is designed to facilitate transactions. It's not designed to preserve the brand-customer relationship in the way that a brand's own digital channel does. That's not a design flaw; it's the product being built. But it creates a real question for retailers about what they're trading away in exchange for the distribution.

The Data Dependency Concern

A third concern that comes up consistently: retailers who build their AI capability on top of third-party AI platforms are creating a dependency on those platforms' continued availability, pricing, and terms of service. This isn't a new concern. The same argument applies to Amazon marketplace sellers, Shopify merchants, and anyone who built significant business on Google search traffic. But the scale and speed at which AI platform dependencies are being created is higher than previous waves.

Amazon, Meta, Microsoft, Salesforce, and Stripe all joined the Universal Commerce Protocol Tech Council in April 2026. These are not companies whose interests are perfectly aligned with the retailers they serve. The terms on which AI platforms will continue to enable commerce — and what they'll charge for it, and what data they'll share — are not guaranteed to remain at their current (largely promotional) levels indefinitely.

The Legitimate Pushback vs. The Noise

To be fair, some of the retailer resistance to AI isn't structural. It's the predictable friction of organisations being asked to change how they work, with ROI timelines that feel long and failure modes that feel risky. That's real too, but it's a different kind of pushback from the structural concerns above.

The eMarketer framing that "most retailers see AI as the growth engine but are stuck in first gear" is probably the more accurate characterisation of where the median large retailer is. Not resistant so much as slow, resource-constrained, and uncertain about which bets to place.

The structural concerns are different, and they're coming from the retailers who understand the technology best. You have to understand it properly to see the margin and relationship implications. Dismissing that as mere conservatism would be a mistake.

The retailers who work out the right answer first will have a structural advantage. Those who don't may find they've traded a customer relationship for a distribution channel. Those are not the same thing.