The Trust Layer Nobody Is Building
The retail industry spent last week debating AI. The more important conversation was about trust — and almost nobody connected it to their stack.
At Shoptalk this year, the most-cited theme wasn't a product announcement or a platform demo. Per Shoptalk's own head of content, the question that kept surfacing across sessions—on stage and off—was deceptively simple: in an age of AI-generated recommendations and agents making purchasing decisions on behalf of consumers, who do people actually trust?
It came up in the Dutch Bros CEO's main stage remarks, where she described her company's relationship with customers in terms of emotion rather than product. It came up when Reddit's CEO noted that 40% of conversations on the platform involve some kind of purchase or shopping decision— not because Reddit is a commerce platform, but because people still seek out human perspectives when they're deciding what to buy. It came up in a session on physical retail where Glossier's store design was specifically called out for placing try-on tables without mirrors, nudging customers to turn to strangers next to them for a reaction rather than relying on their own reflection.
These examples came from very different companies with very different business models. But they were pointing at the same dynamic, one that has a direct and underexplored implication for how organizations should be thinking about their commerce architecture.
The Validation Loop AI Can't Close
Research presented at the event offered a useful data point: when shoppers receive an AI product recommendation, most of them then go and conduct their own research outside of AI. They look for creators who've used the product. They check Reddit threads. They look for signals from people who have no commercial interest in the outcome.
This isn't a sign that AI-driven commerce is failing, but it is a description of how the buying journey actually works at this moment in time. AI is increasingly effective at narrowing the consideration set: surfacing relevant options, filtering by preference, handling the early stages of discovery. But something else closes the sale. A peer review or a creator walkthrough, something that provides customers with a sense that real people with real experience have validated what the algorithm surfaced.
The implication for agentic commerce is worth sitting with. An agent can find your product, and it can make a compelling case for it. But if a consumer's next move is to go validate that recommendation somewhere else—and the evidence suggests they will—then the question of whether your brand shows up credibly in those validation spaces becomes a meaningful part of your commerce performance, not a separate marketing concern.
Where Architecture Enters the Picture
Most organizations treat trust as something earned through brand work and expressed through creative. That framing isn't wrong, but it's incomplete, and in the context of agentic commerce, it misses a category of problem that architecture teams are actually better positioned to solve than marketing teams.
Consider what the validation loop actually requires at a systems level. User-generated content, creator reviews, community discussion, peer ratings… These are the signals consumers are turning to when they step outside the AI interface to confirm a purchase decision. For those signals to do their job, they need to be current, structured, and accessible. They need to exist somewhere that agents and AI interfaces can reach them, not just in a format that renders on a page. They need to be connected to your product catalog in a way that makes them attributable and actionable, not siloed inside a campaign CMS with no meaningful API surface.
When those conditions aren't met, the problem isn't just that your brand looks less credible in validation spaces, it's that your commerce stack is structurally unable to participate in the part of the buying journey where final decisions get made. The agentic layer can do its job, while the trust layer—because it was never really built as a layer—quietly fails.
The Gap the Industry Isn't Naming
There's been considerable and useful work done on what agentic commerce requires at the execution level: machine-readable APIs, universal cart capabilities, standardized product data, and governance frameworks for autonomous decision-making. The infrastructure conversation is maturing. Shoptalk's sessions made clear that the industry understands the operational requirements of the agentic transition reasonably well.
What's less visible is the parallel conversation about trust infrastructure: the systems and data structures that enable social proof, creator content, and peer validation to function as a connected layer in the commerce stack rather than as disconnected assets managed by different teams in different tools.
The organizations at Shoptalk that seemed clearest on this weren't framing it as a technology problem. Lowe's described leaning into its creator network as a way to build credibility with younger consumers— not as a media strategy, but as a structural response to how that audience validates purchase decisions. Macy's framed its AI investment around extending the advisory trust of in-store colleagues into digital channels, treating the human judgment of store associates as something worth preserving and scaling rather than replacing. In each case, the underlying question was the same: how do you build and maintain the signals that give consumers confidence, in an environment where the interface is increasingly automated and the brand's ability to show up directly is more constrained?
That's an architecture question as much as it is a brand one. And most commerce stacks aren't built to answer it.
What This Means Practically
The practical implication isn't that organizations need to rebuild their stacks around trust; it's that trust signals need to be treated as structured, retrievable data rather than as presentation-layer content.
Reviews and ratings that exist only to render on a PDP are not accessible to the agents and interfaces that consumers will increasingly use to validate purchases. Creator content that lives in a campaign tool with no connection to the product catalog is invisible at the moment it would be most useful. UGC that's locked inside a platform CMS can't be served by a third-party AI interface, no matter how good the content is.
Getting this right doesn't require a separate initiative from the agentic infrastructure work already underway. In many cases it's an extension of it— a recognition that the data layer agents need to execute transactions is related to, but not the same as, the data layer consumers need to feel confident making them. Both matter. And for most organizations, the second one hasn't been seriously mapped yet.
The agentic commerce conversation is well underway, and the infrastructure requirements are becoming clearer. What Shoptalk surfaced—in a way that didn't get fully absorbed into the technical discussion—is that execution capability and trust infrastructure are distinct problems, and progress on one doesn't automatically produce progress on the other. Retailers and brands building toward an agentic future would do well to treat them that way.
Leigh Bryant
Editorial Director, Composable.com
Leigh Bryant is a seasoned content and brand strategist with over a decade of experience in digital storytelling. Starting in retail before shifting to the technology space, she has spent the past ten years crafting compelling narratives as a writer, editor, and strategist.