Shopify’s Fall ’26 release made agentic commerce considerably more concrete. Its Shop skill allows AI agents to search more than a billion products, use a shopper’s preferences and recent order history to make recommendations, and complete checkout with Shop Pay. The shopping interface no longer has to be the merchant’s website or app.
But this shift does not automatically cost the merchant the customer. Shopify says an order originating in an AI channel still reaches the merchant with the buyer’s name, email, shipping details, and normal order record. When a shopper connects a Shop account, returning-customer history and preferences can also remain available.
The more consequential change can happen before the order exists.
A large part of ecommerce has been built around observing the path to purchase. Search terms, product-page visits, filters, comparisons, abandoned carts, campaign referrals, and repeat browsing all reveal something about intent. When an AI agent handles more of the searching and evaluation elsewhere, the transaction can arrive with a recognizable customer attached while parts of the reasoning that produced it remain outside the merchant’s field of view.
The trail changes before the transaction
|
Stage |
In a merchant-owned journey |
In an agent-mediated journey |
What may change |
|
Discovery |
The shopper searches, browses categories, follows campaigns, or lands on product pages |
An AI service can search across merchants and assemble options before the shopper visits a store |
The merchant may see less of how the product entered consideration |
|
Evaluation |
Product views, filters, reviews, comparisons, and return visits create behavioral signals |
The agent can narrow choices using information gathered across multiple sources |
Some preference-building activity happens outside merchant analytics |
|
Purchase |
Cart and checkout connect the journey to an order |
The agent or commerce platform can initiate checkout on the shopper’s behalf |
Customer and order data can survive while earlier influence remains harder to attribute |
|
After purchase |
Loyalty, service, fulfillment, returns, and repeat shopping continue through merchant systems |
Many of those interactions can still remain directly connected to the merchant |
The post-purchase relationship may become more important as an owned source of customer context |
The pattern is concentrated at the top of the journey.
McKinsey’s 2026 agentic-commerce research found that 56% of surveyed consumers had used a stand-alone AI chat or search tool in the previous three months, compared with 27% who had used a retailer’s own AI assistant. That gap narrows as the shopper approaches purchase and effectively disappears at checkout. Consumers are therefore more likely to use external AI while deciding what deserves consideration than when the actual transaction is completed.
This changes the information available to the merchant. A retailer accustomed to seeing ten product views, three searches, two returns to the site, and a final conversion may increasingly receive something closer to the end of that sequence. The customer can arrive with a well-formed preference because an agent has already filtered the market. Although the retailer still knows what was bought, it may know considerably less about why that product won.
A distinction like this is significant for personalization systems. A customer record built from purchases alone is much thinner than one informed by what the customer considered, rejected, revisited, or nearly bought. Those signals have traditionally helped shape recommendations and lifecycle communication. If more evaluation takes place inside an external agent, merchants may need to rethink how much they can infer from activity on their own storefront.
The problem is already visible in attribution.
An October 2026 PYMNTS Intelligence study of 60 senior ecommerce decision-makers found that none of the merchants selling through their own websites or apps could fully and consistently connect AI-driven visits to completed purchases. Thirty-seven percent could identify an AI-originated visit but lost the trail before purchase, while another 51% reported only partial or inconsistent attribution. The sample is small and should be treated as directional, but it describes a practical measurement gap rather than a hypothetical one.
It is also possible for the reverse to happen. A merchant can receive the completed purchase and customer identity while missing some of the upstream influence that brought the customer there. Attribution has always had blind spots across devices and channels. But agent-mediated discovery adds another layer because the research environment itself may sit outside the merchant’s analytics stack.
This does not mean direct commerce experiences are fading away. Attentive’s 2026 consumer research found that only 10% of purchasers in the AI-assisted shopping experiences it studied completed the purchase inside AI. Sixty percent said they would prefer to buy through the website or app where the product is sold. Among people who made an AI-assisted purchase, 69% took at least one action to remain connected with the brand or retailer.
The customer relationship can therefore survive agentic discovery; it may simply begin with less observable context.
Payment infrastructure is evolving in the same direction. Visa is developing agentic payment credentials that carry information about who is initiating a transaction, what the agent has permission to do, and the conditions under which the action is authorized. Its newer token capabilities are also designed to carry additional identity and behavioral signals across channels.
The emerging commerce stack is trying to preserve trust and identity as the interface changes. Merchants have a similar problem with customer context.
What still needs to travel with the customer
|
Area |
What the commerce environment needs to preserve |
|
Product truth |
Current product information, availability, pricing, compatibility, and policies that can be understood consistently wherever discovery occurs |
|
Customer continuity |
A reliable way to recognize returning customers, respect consent, and reconnect purchases with loyalty or account history |
|
Transaction context |
Enough information to distinguish agent-mediated activity, understand authorization, and trace an eventual order back to its source where the ecosystem allows it |
|
Operational continuity |
Orders that still connect cleanly with inventory, fulfillment, service, returns, and the systems responsible for completing the promise made at purchase |
|
Measurement |
Analytics capable of separating what is observable from what is inferred, rather than treating an incomplete customer journey as though it were fully captured |
None of these requirements begins with an AI shopping agent. They are familiar commerce architecture problems.
When more of the journey happens outside the storefront, gaps between commerce systems become harder to absorb. An agent can only work with product, availability, pricing, and policy information it can access reliably, while fragmented customer records make it harder to reconnect a purchase with loyalty or lifecycle history. If checkout also passes too little context into downstream systems, the merchant can end up with a completed order and an incomplete view of how it fits into the wider customer relationship.
For years, ecommerce architecture has been asked to connect the storefront to the rest of the business. Agentic commerce extends that boundary outward. The commerce environment now has to work when part of the buying experience is happening on an interface the merchant did not build. This changes the value of the moments the merchant still owns.
The post-purchase relationship gives merchants another opportunity to rebuild some of that context. Service interactions, loyalty activity, fulfillment history, account data, and stated preferences can reconnect a transaction that began elsewhere with a customer the merchant can recognize and serve over time. Those interactions create continuity even when parts of the original discovery and evaluation journey remain outside the merchant’s view.
The storefront remains important, although it may no longer be the place where every purchase begins.
Agentic commerce can preserve the identifiable customer while moving more of the decision journey outside merchant-owned surfaces. Companies need to now understand precisely which signals continue to reach them, which disappear upstream, and whether their commerce systems can maintain continuity when the path to purchase no longer runs entirely through their own front door.
Continue Exploring
As commerce journeys spread across storefronts, AI interfaces, payment systems, and customer platforms, maintaining continuity depends on how well those systems share context. Fulcrum Digital helps commerce teams connect the platforms, data, and operational systems behind the customer experience so new buying models can be introduced without losing sight of how the wider commerce environment works.