There’s a sentence that sounds slightly ridiculous today but probably won’t in a couple of years: your next customer might not actually visit your website.
They might not see the homepage you spent six months redesigning. They might not scroll through the product photography, read the lovingly crafted About page, click your retargeting ad or sit through the 15% off pop-up that appears precisely 0.8 seconds after they arrive.
They might simply say: “Find me a waterproof jacket under £180 that’ll survive a wet week in Scotland, isn’t made from anything too plasticky and will arrive by Friday.”
And an AI agent will go shopping.
It’ll research the options, compare specifications, check prices, assess reviews, rule out the products that don’t fit the brief and come back with a handful that do. Increasingly, it may also be able to put one in a basket and complete much of the transaction.
That’s the basic promise behind agentic commerce, and unlike quite a lot of AI terminology, this one describes something genuinely useful rather than something we were already doing with an extra adjective attached.
The implications for brands, however, are a lot bigger than making online shopping slightly more convenient. Because ecommerce has always been built around attracting a human being, holding their attention and persuading them to make a choice. What happens when a machine starts making part of that choice for them?
The Customer Journey Is Changing
National Express
For most of the internet era, the customer journey has been messy but broadly understandable.
Somebody sees an advert. They Google the product. They visit a few websites. They read reviews. They ask a friend. They disappear for three weeks. They see another advert. They forget why they were looking in the first place. Then, at 11:43pm on a Sunday night, they inexplicably buy it.
Marketers have spent decades trying to map this behaviour with funnels, journeys, touchpoints and diagrams involving far too many arrows.
AI threatens to make those diagrams even messier.
Consumers are already using generative AI to help with the decision-making part of shopping. McKinsey found that 38% of European consumers were using generative AI tools to research products and services and help decide what to buy. Crucially, its research suggests AI is currently having a much bigger impact on influencing purchases than actually executing them. Full autonomy remains much less common. McKinsey’s European agentic commerce research
The thing is, we’re not suddenly entering a world where everybody hands their credit card to a chatbot and tells it to go wild. But we are entering one where an AI layer increasingly sits between intent and purchase.
It’s not difficult to understand the appeal. Shopping online is extraordinarily convenient right up until you actually have to choose something. Search for “best TV under £1,000” and you’re rewarded with comparison sites, affiliate pages, sponsored results, Reddit threads, YouTube reviews, retailer listings and enough model numbers to make you question how badly you wanted a television in the first place.
An AI shopping agent can potentially compress all of that into a conversation.
That’s why this shift sits neatly alongside a broader change we’ve already explored at Creativepool: brands are increasingly having to think not only about being memorable to people, but about being visible and comprehensible to the systems helping those people make decisions. As we noted in our look at what brands are actually looking for in 2026, AI-driven discovery is already changing what it means to be “findable”.
Agentic commerce simply takes that idea one fairly significant step further.
So, What Is Agentic Commerce?
Federico Duran
At its simplest, agentic commerce is shopping in which an AI system doesn’t just answer questions but carries out parts of the shopping process on your behalf.
There’s a big difference between an AI saying, “Here are five running shoes you might like” and an agent being told, “Find me a pair of size 10 trail shoes under £120, prioritise grip over weight, make sure they’re available for delivery this week and put the best option in my basket.”
The first is recommendation. The second is delegation.
And the technical plumbing required to make that delegation possible is already being built.
Google has spent much of 2026 expanding its infrastructure around agentic commerce. Its Universal Commerce Protocol is designed to let AI agents communicate with retailers across discovery, checkout and post-purchase tasks, while its new Universal Cart can pull products from different retailers into a single AI-assisted shopping experience across services including Search and Gemini. Google’s Universal Cart and agentic shopping update
OpenAI is moving in a similar direction. Its Agentic Commerce Protocol, originally introduced to support purchasing within ChatGPT, has since been expanded into product discovery, allowing shoppers to compare products and receive current product information without conducting the traditional multi-tab pilgrimage around the internet. OpenAI’s product discovery update
Shopify, meanwhile, says AI-driven traffic to Shopify stores grew eightfold year-on-year in the first quarter of 2026, while orders originating from AI-powered searches increased almost thirteenfold. Those are Shopify’s own platform figures rather than a universal measure of consumer behaviour, but they give some indication of how quickly AI for shopping is moving from experiment to actual sales channel. Shopify on agentic commerce
The interesting thing here isn’t any single product launch. Platforms launch things constantly and many of them eventually disappear into the same cupboard as the metaverse strategy.
It’s the direction of travel.
Search, discovery, comparison, recommendation, basket creation, checkout and customer service are gradually being connected into something that looks less like a search engine and more like a digital personal shopper.
How Do AI Shopping Agents Decide What to Buy?
Design Studio UI UX
This is where things get rather more interesting for anybody working in branding, advertising or ecommerce.
A human shopper can be persuaded by all sorts of gloriously irrational things. Packaging. Status. Nostalgia. A celebrity they vaguely like. The colour of the bottle. A funny advert they saw three years ago. The fact their dad always bought the same brand.
A machine is, at least theoretically, less susceptible to the seductive powers of a beautifully art-directed 30-second spot.
If someone tells an AI agent they want “the cheapest reliable dishwasher with strong reviews, low energy use and delivery by Saturday”, it has a fairly explicit set of things to optimise for.
That doesn’t mean brand suddenly becomes irrelevant. Quite the opposite, I’d argue. It means some of the signals that make a brand valuable are going to need to become legible to both people and machines.
AI recommendations can draw on product feeds, availability, pricing, reviews, product descriptions, retailer information and whatever other trustworthy data a platform is able to access. Google is already expanding Merchant Center with new attributes designed specifically for conversational shopping, including richer information about common product questions, accessories and possible substitutes. Google’s agentic commerce tools for retailers
In other words, product data that used to feel like the boring plumbing underneath ecommerce is starting to become part of marketing.
If your beautifully branded £300 product is described online as “Model XG44-BLK”, has three contradictory specification sheets and a product feed that hasn’t been updated since Tuesday, an AI agent may not care how gorgeous the campaign was.
The glamorous end of marketing and the deeply unglamorous end of catalogue management are becoming strangely intimate.
That doesn’t mean creativity loses its value. If anything, it gives distinctiveness another job to do. When thousands of functionally similar products are being compared instantly, having something genuinely recognisable about the brand becomes even more important. We’ve already argued that originality becomes more valuable when the world is drowning in content. The same logic applies when the world is drowning in choice.
The machine might narrow the field, but the human still has to want what remains.
What Happens When the Customer Never Visits Your Website?
Katie Miller
This is probably the bit marketers should be losing slightly more sleep over.
The modern ecommerce website is designed around a fairly obvious assumption: eventually, the customer arrives.
Everything follows from that. You optimise the landing page. You test the button. You rewrite the product copy. You tweak the checkout. You improve the photography. You add reviews. You simplify navigation. You argue for three weeks about whether the CTA should say “Buy Now” or “Shop Now”.
But what if discovery, consideration and part of the transaction increasingly happen somewhere else?
Google is already allowing purchases to take place through AI experiences, and OpenAI has demonstrated shopping journeys where discovery and checkout can happen within ChatGPT. OpenAI’s Instant Checkout announcement
That doesn’t make websites obsolete. Brands still need somewhere to establish identity, tell stories, provide depth, support customers and build direct relationships. Plenty of purchases will also remain far too emotional, complicated or considered to be reduced to a neat automated transaction.
But it does mean the website may stop being the front door for every customer.
For some people, your product feed could become the front door. For others it’ll be an AI answer, a recommendation inside Gemini, a ChatGPT comparison or whatever the next interface happens to be.
This is similar to the anxiety publishers have already experienced with zero-click search. You can technically appear in somebody’s journey without that person ever arriving on property you control.
The ecommerce version could be even more consequential because there’s money at the end of it.
That’s why McKinsey’s estimate that AI agents could mediate between $3 trillion and $5 trillion in global consumer commerce by 2030 deserves attention, even though it should be treated as a forecast rather than a destiny carved into a server rack somewhere. McKinsey’s agentic commerce forecast
The number matters less than the underlying possibility: a substantial chunk of future commerce could be influenced by an intermediary that consumers trust to filter the market before brands ever get a chance to speak to them directly.
Are AI Agents Becoming the New Gatekeepers?
Hasham Hussain
We’ve been here before, in a sense. Google became the gatekeeper to the web. Amazon became the gatekeeper to vast swathes of ecommerce. Facebook and Instagram became gatekeepers to audiences. TikTok became a discovery engine. App stores became the route into mobile behaviour.
Every time this happens, brands gain access to enormous audiences while simultaneously giving up a little control over how those audiences reach them.
AI shopping agents could become another version of the same bargain.
The crucial difference is that they don’t simply organise information. They interpret it.
A search engine can give you ten blue links and broadly leave the decision to you. An AI shopping agent can potentially say: “I looked at the options. These three fit what you asked for. This one is probably best.”
That’s a much more powerful position.
It’s also why it’s probably premature to declare that AI agents will “replace search”. Search itself is already becoming more agentic. Google is weaving conversational AI, product comparison, recommendations and transactional tools directly into Search, rather than politely standing aside while something else takes over. Google’s 2026 commerce strategy
The more realistic shift is from searching for information towards asking a system to resolve an intention.
We might search “hotels Copenhagen June” today. Tomorrow the request is more likely to be: “Find me somewhere central in Copenhagen for three nights in June, under £220 a night, quiet enough to work from, with excellent coffee nearby, and don’t show me anywhere reviewers complain about the beds.”
That sounds like a subtle change in interface but it really isn’t. One gives the shopper information and the other gives the shopper judgement.
What Does Agentic Commerce Mean for Brands?
Jackie McAdam
The temptation here will be to invent a whole new discipline, attach the word “agentic” to it and begin selling £40,000 transformation workshops by Thursday.
Brands probably don’t need that.
They do, however, need to start thinking about what happens when the customer journey has an AI-shaped middleman.
Part of the answer is technical. Product information needs to be accurate, structured, current and detailed enough for AI systems to understand what you actually sell. Inventory, pricing, variants, specifications, reviews, delivery information and returns policies become increasingly important machine-readable signals.
But reducing the challenge to structured data would be a mistake.
Agents still act on behalf of humans, and human preferences aren’t suddenly disappearing because the interface changes. Somebody might ask for “the most reliable washing machine”, but they might just as easily ask for “a beautiful coffee machine that won’t make my kitchen look like an office” or “trainers from a company with decent environmental credentials that don’t look painfully sensible”.
Those requests contain judgement, values, identity and taste.
Brands still have to create those things.
In fact, if agents get better at filtering out meaningless choice, there’s an argument that genuinely distinctive brands become even more valuable. A market full of interchangeable products becomes very easy for an AI system to reduce to price, convenience and specification. If you don’t want to compete purely on those terms, you need to give people reasons to care about things that aren’t quite so easily reduced to a spreadsheet.
That’s one reason why the future of branding isn’t simply about adopting more technology. Brands will need strong identities, trusted reputations, clear positioning and useful data at the same time. Being technically discoverable won’t help much if there’s nothing distinctive to discover.
There’s a creative challenge here too. We recently asked whether brands will still need agencies when AI can make the work. Agentic commerce adds another wrinkle to that question. Agencies won’t just be thinking about what consumers see. They may increasingly have to think about what machines understand, retrieve, compare and repeat about the brands they represent.
That doesn’t sound particularly sexy but neither did SEO in 2004.
So, Is the Machine Really the Customer?
Mustafa Albarbary
Not quite.
Calling AI systems “machine customers” makes for a wonderfully alarming headline, but the human hasn’t disappeared from the equation. They’re still providing the money, setting the preferences, establishing the boundaries and, in most current systems, retaining control over important purchase decisions.
The machine is better understood as an increasingly powerful representative.
And that might actually be the more profound idea.
For years, brands have spent billions trying to understand customers: tracking behaviour, segmenting audiences, analysing purchase histories and predicting intent.
Now customers are beginning to acquire software that understands them too.
An AI agent could potentially know your size, budget, dietary restrictions, favourite brands, hated brands, loyalty schemes, travel plans and tendency to refuse to pay £6.95 for delivery out of principle. It doesn’t have to be persuaded from scratch every time because it’s carrying a version of your preferences around with it.
Suddenly the balance shifts.
Brands aren’t simply trying to understand the consumer anymore. The consumer has an increasingly capable machine sitting on their side of the table.
That could make shopping considerably easier. It could also make lazy marketing much less effective. Artificial scarcity, confusing pricing, deliberately difficult comparisons, terrible product information and websites engineered to wear customers down become a little less powerful when somebody has an infinitely patient robot prepared to do the annoying bits.
And perhaps that’s ultimately where the excitement sits.
The most interesting thing about AI shopping agents isn’t that robots are going to start developing passionate opinions about trainers.
It’s that humans may soon have technology capable of representing their interests in markets where the technological advantage has traditionally belonged almost entirely to the seller.
Your next customer will still be human.
But increasingly, you might have to impress their agent first.
