For two decades, retailers have competed for one thing: a customer's attention. They built websites, tuned search rankings, ran loyalty programmes, and spent heavily on brand. The whole game was making it easy for a person to find you and choose you.
That game is changing. More shoppers will soon start with an AI assistant. They'll ask it to research a product, compare options, and narrow down their choices before they ever land on a retailer's site. In some categories, the assistant will soon handle the purchase itself. So retailers are no longer only competing for attention. They are competing to be the product an agent recommends.
An agent does not shop the way a person does. A customer who finds an incomplete product page will often fill the gap themselves. They read the reviews, open another tab, and message customer service. An agent will not do any of that. If it cannot verify what it needs, it moves to the next option.
The UK is one of the better-prepared markets for this shift. Online already accounts for about 28% of retail sales, among the highest shares in Europe (ONS, early 2026). Open Banking processed 351 million payments in 2025, and Variable Recurring Payments are starting to give agents a trusted way to transact on a customer's behalf. The appetite is there too. A Klaviyo survey in late 2025 found that 80% of UK shoppers already use AI tools when they shop, and 70% expect AI assistants to be a normal part of the buying process soon. But appetite is not the same as a good experience. In a separate CI&T study, 68% of UK and Ireland shoppers could not name one AI shopping experience that genuinely impressed them.
That gap is the opportunity, and it sits where few people are looking. The instinct is to reach for a smarter model or a slicker chatbot. The harder, more useful work is making your business legible to a machine. Can an agent confirm fit and sizing? Is the item actually in stock right now? Can it arrive in time? Are the return terms clear? When those answers are missing or inconsistent, you become the risky choice for the agent, and it quietly picks someone else.
In one of our shopping-agent deployments, on a marketplace app with more than 100 million downloads, a search for a black T-shirt started returning a purple one. The customer was not happy. Troubleshooting showed the catalogue tagging was wrong: the purple shirt had black in its primary colour field. In any agent deployment, the catalogue supersedes whatever logic you train the agent on.
This is why product data is becoming one of the assets that decides who wins. For years most retailers treated it as housekeeping, an operational task to keep tidy. The retailers most likely to benefit from AI are not the ones with the most advanced models. They are the ones with the cleanest catalogues, accurate inventory, and a reliable fulfillment system. With the agents we run for one of the world's top five retailers by store count, we've seen revenue per session rise by as much as 7%.
The cost of waiting is real. Adobe found that traffic arriving from AI sources converted about 31% better than non-AI traffic over the 2025 holiday season, and 42% better by March 2026.
Retailers also lose around £38 billion a year to abandoned baskets (Retail Economics, 2024). Agentic commerce will not fix that overnight, but it can take friction out of the journey. The quieter advantage is learning. Every interaction an agent handles generates data about intent and trust that you cannot buy in later. A year from now, the distance between the retailers who started and the ones who waited may be wider than they expect.
The next step is bigger than shopping help. It is delegated spending. Instead of asking an assistant to place an order, people will ask it to manage a category on their behalf, replenishing the things they buy on a rhythm without being asked each time. Small businesses will let agents handle routine procurement within set limits. The UK is well placed for it. People are comfortable with digital payments, Open Banking keeps maturing, and the rules are catching up.
On the rules, UK regulators have been consistent. The FCA, the ICO, and the DRCF have all made the same point recently: a business stays accountable for the decisions its AI makes. You cannot hand responsibility to an algorithm. So governance and auditability matter as much as technology does.
That points to a sensible order of work. The first step is not launching an AI shopping assistant. It is fixing the foundations: product data, inventory accuracy, and the operational consistency underneath them. Start with high-volume, low-risk jobs like delivery updates, returns, and post-purchase questions, where an agent earns trust without much downside. Move to recommendations, automatic replenishment, and delegated buying after that.
The first era of eCommerce rewarded the retailers a customer could easily find. The next will reward the ones an agent can easily trust. Most of the conversation about agentic commerce is still about models and chatbots. The more useful one, for UK retailers, is about data, operations, and trust. Sort those out, and you become the retailer an agent is comfortable recommending.