Topic
Agentic commerce
Customers are starting to ask rather than browse. The brands building their own agentic surface stay in the conversation. The brands that don't get summarised away.
What is agentic commerce?
Agentic commerce is the practice of selling through AI agents that act on a customer's behalf, rather than through static product catalogues and search interfaces.
The agent might live inside an LLM the customer is using, on a brand's own site, or in a third-party shopping interface. Wherever it lives, it understands intent, reasons over product data, and walks the customer towards a decision through conversation rather than browsing.
Agentic commerce is a new commerce surface, not a new checkout flow. The static storefront does not disappear immediately, but it stops being the first place most customers encounter the brand. The agent does.
For brands, the strategic question is whether that agent belongs to the brand or to someone else.
Why this matters now
Three shifts are landing on top of each other, and brands without a position on agentic commerce will find themselves at a structural disadvantage.
The discovery moment is moving inside LLMs.
Customers are asking ChatGPT, Gemini, Perplexity and Claude what to buy. Google still handles roughly a billion shopping queries a day, but ChatGPT alone now handles 50 million, and the trajectory is steep. The interfaces are designed to keep the conversation inside the model and prioritise engagement with the platform over conversion for the merchant. Walmart's ChatGPT integration converted worse than Walmart's own .com, a pattern that holds across most current LLM-to-merchant integrations.
Customers expect conversation, not navigation.
A list of ten thousand results ranked by ad spend feels increasingly outdated next to asking a question and getting a useful answer. The shift is fastest in high-consideration categories (electronics, furniture, automotive, fashion with detailed specs) but is visible across most ecommerce verticals.
The brand-owned agentic surface is becoming a defensive necessity.
A brand without its own agentic answer to "what should I buy" ends up wholesaling to whichever LLM the customer asked first. The asymmetry favours the LLM permanently if the brand stays passive. The brands building their own agentic surfaces today are building a structural defence against that.
The component parts
Agentic commerce is built from three architectures that brands typically buy separately.
Brand-owned product data
Structured, verified, machine-readable data about each product. Materials, specs, compatibility, repairability, sustainability claims. Required for compliance under Digital Product Passports for regulated categories from 2027, useful far beyond compliance. The richer the data, the more credibly an agent can answer questions. See our Digital Product Passports page for the deeper read on the data layer.
A conversational and agentic surface
The data has no commercial value if customers cannot reach it conversationally. A brand-owned agent that lives on the brand's own site, in store, or wherever the customer is. Our product TalkPod is the working version of this at scale across automotive, property, retail, technical products, and B2B.
An MCP or API layer that lets external agents query the brand's data on the customer's behalf
This is the bit most brands have not started thinking about. When the customer asks an LLM "should I buy this", the LLM is going to look for an answer somewhere. Brands that expose their data through an open protocol stay in the conversation. Brands that do not are summarised from whatever the LLM has scraped.
The brands that win the next decade will be the ones that treat these three as one architecture instead of three vendor decisions.
What it looks like in practice
Four concrete examples of agentic commerce at work.
A customer asks the brand's conversational agent which jacket has the longest repairability window for use in wet UK conditions, and gets a verified answer drawn from the products' own records, not a marketing claim. The brand wins the moment of evaluation in a way it never could with a comparison grid.
A trade buyer at a counter scans a drill and asks the pod whether its battery is compatible with the cordless tools they already have at home. The pod answers in seconds, drawing on the brand's structured product data. The buyer walks out with the drill. The brand learns that compatibility is the question that converts.
A repeat customer asks an LLM for a holiday outfit recommendation. The LLM queries the brand's product data through an open API, gets a real answer, and links back to the brand's own checkout. The brand stays in the customer relationship even though the discovery moment happened inside the LLM.
A retail leadership team installs a new conversational layer on their .com and quietly retires the dropdown filter UI nobody uses any more. Conversion rates lift, average order value lifts more, and the team starts seeing intent data their analytics never captured before.
None of these require a new physical product. What they require is the agentic commerce architecture built in advance.
Frequently asked questions
What is agentic commerce?
Agentic commerce is the practice of selling through AI agents that act on a customer's behalf, rather than through static product catalogues and search interfaces. It covers brand-owned conversational surfaces, integrations with third-party AI shopping agents, and the data layer underneath both. The shift is from browsing and filtering to asking and being answered.
Is this the same as a chatbot?
No. A chatbot answers questions inside a single channel from a fixed script or a generic model. An agentic commerce surface runs on the brand's structured product data, reasons over it conversationally, and can act on the customer's behalf (recommending, comparing, completing a purchase). The infrastructure underneath is closer to a product database with reasoning over it than to a customer service bot.
Will commerce platforms eventually offer this as a feature?
Probably yes, on a timeline measured in months not years. The reason to build now isn't competing with the platform's eventual feature; it's owning the brand-specific data layer, the agent's voice, and the customer relationship before the platform default is good enough to be the answer. Brands that wait until the platform ships the feature will have generic agents indistinguishable from competitors using the same defaults.
How does this relate to Digital Product Passports?
DPPs provide the verified, structured product data layer that an agent reasons over. They are mandatory under EU regulation for batteries from February 2027 and most consumer categories through 2030. The brands that approach DPP only as compliance miss the commercial opportunity: the same data that satisfies the regulator powers the agent that answers the customer's questions.
What is the role of LLMs like ChatGPT and Claude?
LLMs are increasingly where customers ask their first question about a product. The commercial question is whether they conclude the conversation inside the LLM (the LLM keeps the customer, the brand gets summarised away) or whether they can query the brand's own data and route the customer back to the brand for the decision and the purchase. The Model Context Protocol is one open standard that lets brands expose their data to LLMs deliberately.
Where do brands typically start?
Brands tend to start with one of three entry points. A conversational layer over existing product data on the brand's own site. A DPP pilot on a single product line. Or an MCP or API integration that lets external agents query the brand's product data. The right starting point depends on which conversation the brand can have credibly today and which regulatory or competitive pressure is closest.
How we help
Where agentic commerce lives in our work
We work across the data layer, the conversational surface and the strategic programme as one architecture.
TalkPod
Our conversational product platform. We connect TalkPod to your existing product data, configure it for your range and customers, and run a working pilot in one channel. You see what it does, what it changes, and what scaling it would look like.
Explore TalkPodDigital Product & AI
Product strategy, service design, and practical AI integration. We help you decide what to build, where AI genuinely adds value, and how to wire the agentic surface into the rest of the product.
View serviceDigital Strategy & Delivery
Senior digital leadership for the broader programme. We embed in your team to take the agentic commerce work from strategy through to operational use across product, data, and customer experience.
View serviceRelated thinking
Insight
Are agentic stores the future?
Swap launched the first dedicated agentic storefront recently, an early outlier that raises many strategic questions for brands and how they are accessed online.
Opinion
When the browser becomes the destination
Most websites weren't designed to be queried. They were designed to be visited. That distinction is about to matter more than most organisations realise.
Insight
When Products Can Answer Questions
Digital Product Passports start rolling out in 2027, attaching structured, verified data to millions of physical products. At the same time, the way people search for and evaluate those products is shifting, moving away from browsing and filtering towards asking questions and getting direct answers.