Designing retail for AI agents as well as people

A lot of energy has been spent refining the online shopping experience for people. Websites have become more visual, product pages more immersive and checkout journeys increasingly seamless, all designed to guide a human shopper from discovery to purchase with as little friction as possible. The industry has become very good at understanding how people browse, compare and ultimately decide what to buy. UX is big business.

 

But as always, nothing stays the same. Now another type of customer is beginning to enter that journey. Not a new demographic or another sales channel, but something fundamentally different. A machine.

 

Agentic commerce, where AI assistants search for products and eventually complete purchases on behalf of users, is quickly moving from theory into practical reality. Consumers are already asking AI tools to find the best running shoes under a certain price, a gift for a specific person, or a dress that matches something they already own. And the fact OpenAI is stepping out of the field doesn’t suggest the market is waning - it opens it up for more. The logical next step is simple. Instead of presenting a shortlist of options, the assistant places the order. This development introduces a new challenge for retailers because the commerce experience now has to work for two very different audiences at the same time. Humans and machines.

 

Human shoppers behave in ways retailers understand well because the industry has spent years studying and designing around those behaviours. We know that people browse, explore, scroll through multiple pages, compare colours and sizes, read reviews and often change their mind halfway through the journey. They respond to imagery, brand storytelling and merchandising cues that help them imagine the product in their lives. Emotion and psychology plays a huge role in purchase, often much more than rational though. Retail design has evolved to support exactly this kind of behaviour, combining visual inspiration with practical information to move someone gradually towards a purchase.

 

AI agents approach the same task in a completely different way. They are not browsing for inspiration or responding to visual storytelling. Instead, they are looking for structured information they can interpret quickly and reliably, extracting clear signals that help them make a decision on behalf of a user. That means they seek accurate product attributes, reliable pricing, up to date inventory, clear delivery timelines and transparent return conditions. If an AI assistant is going to recommend a product or place an order with confidence, the information it relies on must be consistent and accessible in real time.

 

This is where the underlying challenge for retailers becomes clear. Their infrastructure now needs to speak two languages at once. One optimised for human experience and one optimised for machine interpretation. For a person, a product page full of images, styling advice and brand storytelling makes perfect sense because it provides context and emotional cues that help shape a decision. For an AI agent, that same page can be surprisingly difficult to interpret if the underlying data is not clearly structured and synchronised across systems.

 

And for many, it’s not so easy. Behind the polished front end of most e-commerce websites sits a complex network of operational technology. Product data might live in one platform, pricing in another, inventory within a warehouse management system and fulfilment information somewhere else entirely. Customer data, promotions and logistics are often spread across yet more systems. There are multiple marketplaces involved, and physical locations that span geographies. We work with retailers who may have a hundred different operations happening in the background every time an order is placed. To a human shopper, this complexity is largely invisible and rarely causes immediate confusion. If a size is unavailable or a delivery option changes, people can adapt their choices and continue browsing. They might select a different product, choose a slower delivery option or return to the search results.

 

AI agents operate very differently because they rely entirely on the accuracy of the information provided to them. If the data is incomplete, inconsistent or delayed, the agent simply cannot complete the task reliably. Imagine a customer asking an AI assistant to find black running shoes, size nine, under £120, with delivery by the weekend. The assistant scans several retailers and identifies a pair that appears to meet those criteria. When it attempts to place the order, it discovers the size is actually out of stock or that the delivery estimate is inaccurate because the inventory data was not synchronised across systems. From the consumer’s perspective, the AI assistant appears unreliable. In reality, the problem sits deeper in the retailer’s operational architecture.

 

Large language models and AI assistants are remarkably good at exposing these hidden gaps. As soon as machines begin interacting directly with retail systems, they reveal where product information, pricing and inventory data are fragmented or inconsistent. This is why the real conversation about agentic commerce should not begin with the design of the front-end interface. It needs to start with the operational technology stack that sits behind it. When it doesn’t go right, it is costly - our research shows 60% of UK retailers have lost revenue due to poor integrations and connections in the background.

 

If AI agents are going to act on behalf of customers, retailers must ensure that product, pricing and inventory information flows cleanly across their entire commerce ecosystem. E-commerce platforms, ERP systems, warehouse management tools, logistics providers and customer databases all need to remain synchronised so that the information available at the moment of purchase is accurate and reliable.

 

This is precisely where integration platforms play a vital role. Their focus is on connecting the different systems that power modern retail operations so that data moves consistently and in real time across the entire stack.

 

In the context of agentic commerce, that role becomes even more important because AI agents do not experience the front end of a website in the same way a human does. Instead, they rely on the structured data flowing through those systems. If that information is not aligned, the agent cannot make trustworthy decisions.

 

Retailers therefore face a subtle but important shift in thinking. Their technology infrastructure is no longer just supporting the human customer experience. It is also supporting machine decision making. And that can be complex and confusing.

 

The brands that succeed in this new environment will be those that treat AI agents as another type of customer interacting with their systems. That means ensuring product information is structured and consistent, inventory updates appear instantly across channels, pricing remains synchronised, and delivery promises are supported by accurate fulfilment data.

 

When those foundations are in place, retailers can support both audiences effectively. Humans exploring rich, visual storefronts and AI agents making rapid, data-driven decisions on behalf of the people they represent. The front end of retail will continue to evolve as new interfaces and assistants emerge, but behind every one of those experiences lies the same requirement. Clean, connected and reliable operational data. In the age of agentic commerce, good UX is no longer just about what human shoppers see on the screen. It is about whether the systems behind that screen can communicate clearly with machines as well. It might not seem as sexy; it is essential.

 

---

 

Jim Herbert is CEO of Patchworks

 

Main image courtesy of iStockPhoto.com and WANAN YOSSINGKUM