The next person to shop your store may not be a person at all
Agentic commerce is reshaping how North Americans buy. Retailers whose product data is not ready for AI shopping agents will lose sales they never even see.
For thirty years, online retail has run on one ritual. A shopper types a few words into a search box, scans a page of links or product tiles, clicks, compares, and eventually buys. That ritual is ending. The shopper of 2027 is far more likely to ask an assistant to “find me a waterproof jacket for a weekend in the Rockies under 150 dollars” and let the software do the comparing. The search box is being replaced by a conversation, and the consequences for retailers across the United States and Canada are bigger than most boardrooms have grasped.
This is agentic commerce: AI assistants that discover, compare, and increasingly buy on a customer’s behalf. It is not a thought experiment. Grand View Research values the agentic commerce market at 5.7 billion dollars in 2025, rising to 65.5 billion by 2033. McKinsey estimates agentic commerce could orchestrate up to a trillion dollars of US retail revenue by 2030, and three to five trillion globally. Bain puts the US market at 300 to 500 billion dollars by 2030, a quarter of all e-commerce, while Morgan Stanley models a 385 billion dollar impact on the bull case. The numbers differ. The direction does not.
The shift is not only consumer-facing. Gartner forecasts that AI agents will intermediate more than 15 trillion dollars of B2B purchasing by 2028. And the behaviour is already here: Bain reports that between 30 and 45 per cent of US consumers now use generative AI to research products and compare options before they buy. The same race for an enterprise-grade infrastructure layer for AI agents is now playing out across North American venture funds.
The plumbing is being laid faster than the strategy
OpenAI and Stripe launched instant checkout inside ChatGPT in late 2025, and Google, Visa, and Mastercard have all shipped protocols and payment rails that let an agent transact on a shopper’s behalf. When the people building that infrastructure describe it, they are blunt. Stripe’s president and co-founder John Collison told Bloomberg’s Odd Lots podcast this year that “keyword search is ridiculous”, arguing it made sense for buying a book whose title you already knew, and little else.
Here is what should concentrate the minds of retailers and their investors. In an agent-led journey, the recommendation happens before the click. A shopper asks, the assistant weighs the options it can understand, and it puts forward the one it trusts. If your product data is thin, inconsistent, or written for a 2010 search engine, the agent does not argue with you. It recommends the competitor it can read. You never see the lost sale, because it was never a visit in the first place.
From SEO to recommendation visibility
For a decade, visibility meant ranking on Google. The new equivalent is recommendation visibility: whether an AI system can understand who your product is for, what problem it solves, and why it beats the alternative. Most retailer feeds answer none of those questions. They were built for keywords, categories, and cost-per-click bidding, not for machines that reason over structured data. That gap between what retailers publish and what agents need is the single biggest commercial risk, and opportunity, in retail right now. Canadian AI leaders like Cohere are already betting their roadmaps on data sovereignty and enterprise reasoning over consumer chatbots, and the same logic now applies to the commerce stack.
North American businesses are particularly exposed, and not for lack of ambition. Main Street merchants and big-box chains alike have absorbed one structural shock after another, from the shift online to supply pressure and thinning margins, and the instinct after each has been to move cautiously. Caution is the wrong response here. Retailers who structure their data early will compound an advantage while rivals are still debating whether any of this is real. Those who wait will find that being invisible to an agent is far harder to reverse than slipping a few places in a search ranking.
The cost of entry is low — the cost of waiting is not
The good news is that the cost of entry is low and falling. Getting into the major AI shopping channels is not a matter of buying placement. OpenAI runs a free merchant feed program: you verify your business and submit a structured product feed, and Shopify and Etsy catalogs are already wired in. ChatGPT’s shopping results are not advertisements. Visibility is earned through clean, complete, trustworthy data, which leaves the field unusually open for smaller merchants willing to do the work. It is also why a new layer of specialists has emerged to do that work for retailers. Vendoora structures and enriches product and service data so AI agents can understand and recommend it. The wider point holds regardless of who does it: the data has to be ready before the agent arrives. The same generation of Canadian AI startups winning in adjacent verticals proves how quickly the playing field can re-form once the infrastructure is genuinely in place.
None of this removes the human. People will still form brand preferences, still care about trust, still make the final call on anything that matters. But the moment of comparison, the point at which one product is chosen over three others, is moving inside the machine. The brands that win the next decade of retail will be the ones an AI can understand, compare, and confidently recommend.
The shop window is being rebuilt, and this time it is made of data rather than glass. Retailers who treat that as a technical afterthought will pay for it in sales they never knew they lost. Those who treat it as strategy will own the recommendation, and with it the customer.