Agentic Commerce Has Arrived: When AI Agents Shop, Compare And Check Out For Your Customers, Can They Still Buy From You?

72% - Of shoppers now use AI assistants (ChatGPT, Claude, Gemini) for ideas and recommendations  ·  Catalogue to checkout - Shopping and merchant agents now connect to product catalogues, carts, checkout systems, preferences and purchase histories  ·  Agent-mediated - The buyer doing the finding, comparing and transacting is increasingly an AI agent, not a person on your site  ·  Integration challenge - Staying buyable in an agent world is as much an integration problem as a marketing one

A quiet but profound change is underway in how buying actually happens. For the whole history of e-commerce, the model was a person visiting your website, browsing, comparing, and clicking buy. Increasingly in 2026, that person is being replaced in parts of the journey - and sometimes all of it - by an AI agent acting on their behalf: researching options, comparing them, and in a growing number of cases completing the purchase. This is no longer a futurist's slide. Some 72% of shoppers now use AI assistants like ChatGPT, Claude and Gemini for shopping ideas and recommendations, and the infrastructure for agents to actually transact is arriving fast - Anthropic, for instance, has released ready-made templates for Claude-based shopping and merchant agents designed to connect directly to product catalogues, carts, checkout systems, preference databases and purchase histories across retail, travel, telecom and ticketing. The name for this is agentic commerce: buying mediated by AI agents rather than done by hand. And it poses a new and fundamental question to every business that sells anything: when an AI agent is doing the finding, comparing and buying, is your product something it can actually discover, understand and transact with? As an AI Agency London focused on integration, we think this is one of the most important shifts for businesses to grasp, and this guide is how to think about it.

Why This Is An Integration Problem, Not Just A Marketing One

It is tempting to treat agentic commerce as purely a marketing challenge - being recommended by the AI - and that half matters, but the deeper and more overlooked half is integration. Being recommended is worthless if the agent then cannot get accurate information about your product or actually transact with you. Think about what a shopping agent needs to do its job on your behalf: it needs to discover that your product exists and is relevant, which requires your catalogue and content to be machine-readable and clearly described; it needs current, accurate information about specifications, price and availability, which requires that data to be structured and exposed rather than buried in images or inconsistent pages; and, at the sharp end, it may need to connect to your cart and checkout to complete a purchase, which requires the technical plumbing to let an agent transact - the catalogues, carts, checkout systems and preference data that the new merchant-agent templates are built to connect to. Each of these is an integration and data problem, not a slogan. A business whose product data is messy, whose availability is not exposed cleanly, and whose checkout an agent cannot connect to will simply be skipped - not because it was outbid on marketing, but because it was technically unbuyable by the agent. This is why agentic commerce belongs as much to your systems and data teams as to your marketing team, and why the businesses that win it will be the ones that treat being connectable to agents as core infrastructure.

• Discoverable - your catalogue and product content must be machine-readable and clearly described so an agent can find and understand what you sell.
• Accurate and structured - specifications, price and availability need to be exposed as clean structured data, not buried in images or inconsistent pages.
• Transactable - increasingly an agent needs to connect to your cart and checkout to complete a purchase; if it can't, you're skipped.
• Current - agents act on live data; stale prices or availability lead to failed or wrong transactions and lost trust.
• It's systems work, not just marketing - being buyable by agents is an integration and data problem your technical team owns.

What UK Businesses Should Do Now

The practical response for a UK business is to start making itself legible and connectable to agents before agent-mediated buying becomes the majority of the journey, because the groundwork takes time and the early movers will be the ones agents can actually transact with. The first step is to get your product data in order: accurate, consistent, structured, machine-readable information about what you sell, its specifications, price and availability - the foundation everything else rests on, and something many businesses have neglected because human shoppers tolerated messiness that agents will not. The second is to make sure your business is discoverable and clearly described across the web so that agents (and the AI assistants recommending them) can identify what you do and who you serve - the same generative-engine-optimisation discipline that governs being recommended. The third, and more forward-looking, is to prepare your commerce infrastructure to connect to agents - to understand the emerging standards and templates for merchant and shopping agents, and to be ready to let a trusted agent access your catalogue, availability and checkout, so that when a customer's agent comes to buy, it can. None of this requires abandoning your human-facing store, which will matter for a long time yet; it means adding the machine-facing layer that agentic commerce demands. The businesses that build that layer early will remain buyable as the buyer changes from a person to an agent - and that combination of clean data, discoverability and agent-ready checkout is exactly the integration work we do.

The Bottom Line

Agentic commerce - buying in which an AI agent researches, compares and increasingly transacts on a customer's behalf - has moved from speculation to reality in 2026, driven by shoppers trusting AI assistants at scale and by merchant-agent infrastructure that connects agents to catalogues, carts, checkout, preferences and purchase histories. It inserts a new intermediary between your business and your customer, and that intermediary has requirements a human shopper never imposed: your product must be machine-discoverable, accurately and structurally described, current, and technically transactable. Crucially, this makes agentic commerce as much an integration and data challenge as a marketing one - being recommended is worthless if an agent cannot then get accurate information or complete a purchase with you. The response for UK businesses is to build the machine-facing layer now: get product data clean, structured and exposed; make the business discoverable and clearly described so agents can find and understand it; and prepare commerce infrastructure to connect to trusted agents at checkout. The human-facing store still matters and will for years, but the businesses that add the agent-ready layer early will stay buyable as the buyer shifts from a person to an agent - and those that do not will be quietly skipped, not outbid but simply unreachable. Making your business legible and connectable to the agents now doing the buying is the integration work that decides whether you remain in the running at all.

References & Further Reading

• MarketingProfs - AI Update, September 11 2026 (Anthropic shopping and merchant agent templates): https://www.marketingprofs.com/opinions/2026/55845/ai-update-september-11-2026-ai-news-and-views-from-the-past-week
• Optimove - 2026 holiday consumer report (72% of shoppers use AI assistants for recommendations): https://www.optimove.com/resources
• Anthropic - building agents for commerce (shopping and merchant agents): https://www.anthropic.com/
• Trend Hunter - top retail trends, September 2026: from AI shopping assistants to pop-ups: https://www.trendhunter.com/slideshow/september-2026-retail
• Agile Brand Guide - marketing technology and AI news, September 4 2026: https://agilebrandguide.com/yesterdays-marketing-technology-ai-news-september-4-2026/