When the storefront becomes the agent

Swap launched what it is calling the first agentic storefront earlier this month. The product is early, but the category it sits inside is real, and it forces a strategic question every product brand will need to answer in the next twelve months.

The pitch is straightforward. Brands have spent the last twenty years optimising a website that customers find through Google, navigate through a familiar grid, and convert through a checkout funnel that’s been A/B tested into something close to a mathematical limit. That whole pattern is starting to come apart at the front end, because customers are no longer arriving at the site the way they used to. They are asking ChatGPT, Gemini, Perplexity and Claude what to buy, and those interfaces are designed to keep the conversation inside the model rather than to send the customer anywhere.

Swap’s argument is that this is structurally misaligned with what brands actually want, and on that point they are right. The LLMs have no commercial reason to send a visitor to a brand’s site. They have a strong commercial reason to keep the visitor inside the LLM.

The data point worth noticing is that even when the LLMs win the discovery moment, the conversion moment still belongs to the brand if the brand has anywhere worth sending the customer. Walmart’s own testing of around 200,000 items found that conversions inside ChatGPT ran roughly three times lower than clickthroughs to walmart.com, a comparison Swap itself acknowledges. Sephora’s ChatGPT app and Gap’s Gemini partnership are both public but their conversion numbers aren’t.

What Swap have built is a dedicated AI agent that lives on an AI sister site separate from the brand’s .com. It runs on the brand’s product data, sounds like the brand, and walks a customer through a guided journey from discovery to virtual try-on to chat or voice checkout. They are claiming it converts at twice the industry rate, which is the kind of headline number worth reading with care. Industry rate of what, against which baseline, with which traffic mix, on which categories of product. Named launch partners include SIMKHAI, Retrofête, Odd Muse, Studio Nicholson and Manors Golf, alongside 20+ other brands. Small, design-led, high-intent. Not Walmart.

But the category is real because the underlying change is real. We have written about it on this site as the move from visit-based design to agent-led discovery. Customers are starting to ask rather than to browse, and they expect a useful answer rather than a list of ten thousand results ranked by ad spend. A brand that doesn’t have its own agentic surface within the next few years will end up wholesaling to whichever LLM the customer asks first.

The more interesting question for a brand thinking about this seriously is not whether to buy Swap. It is whether the brand has a single coherent answer for its agentic surface across the whole customer journey, or whether it is about to spend the next three years buying separate AI vendors for discovery, product information, post-purchase, service, repair and resale.

The agentic storefront sits naturally next to the agentic product passport, which sits naturally next to the agentic service layer. They are all the same architecture pointing at different moments in the customer relationship. Today, the default is to buy them in silos, with the same disconnection that already exists between IoT and DPP programmes. We have written about that pattern too. The brands that win the next decade will be the ones that recognise their AI surface is one surface, not five, and design it accordingly.

Swap is a useful artefact for that conversation. The strategic question it forces every retail leadership team to answer this quarter is the right one. Where will your customer find you when they stop typing into a search box and start asking an agent, and what do they meet when they arrive?

If the honest answer is “they meet our regular .com and hope for the best”, it is later than you think.

Brand-owned AI surfaces, from discovery through to end of life, are at the core of what we have been building towards at Human Kind. If you’re working through how the pieces fit together, we should talk.