Three Futures for Brand-Led Retailers in an Agentic World

AI agents are no longer an idea in retail. They are a reality. The future of agentic commerce is less about the agents themselves and more about the environment forming around them.

For years, retail technology has added new layers between brands and demand. Search created one layer. Marketplaces created another. Social platforms created another. Each one arrived with a clear benefit: more reach, more efficiency, more convenience, more data. Each also changed the terms of access. Retailers gained new ways to find customers, but often through environments they did not own.

Agentic commerce is forming the next layer. The customer may not begin with a search bar, a category page, or a brand site. They may begin with a request: find this, compare that, choose the best option, complete the purchase. The agent becomes the first interface. That changes the structure of commerce before it changes the transaction itself.

The issue is not that agents will exist. They will. The issue is who owns the environment in which they operate, who sets the rules for participation, and who benefits from the value created when a customer chooses to buy.

From here, three futures are starting to take shape.

Future One: Ecosystem Lock-In

In the first future, agentic commerce consolidates inside a small number of large platforms. Retailers participate because the customer demand is there, but participation comes with a familiar condition. The brand has to fit the platform’s model of commerce.

This means product data must be structured according to the platform’s requirements. Content must be written so the agent can understand and surface it. Assortment, availability, reviews, pricing, fulfillment promises, and service policies all become inputs into a system the retailer does not control. The retailer can optimize for that system, but it cannot govern it.

This is the ecosystem lock-in scenario. The cost is not always visible as a line item. It appears gradually as dependency. The retailer becomes better at operating inside the platform, while the platform becomes more important to the retailer’s access to demand. Over time, the retailer adapts to the platform, while the platform becomes increasingly difficult to replace.

Amazon is the version of this pattern retailers know best. A brand selling on Amazon owns its products and its margin, but not the relationship. The customer believes they bought from Amazon because, in every meaningful sense, they did. Amazon owns the search, the comparison, the recommendation, the checkout, the return, and often the next purchase. Leaving the ecosystem becomes structurally painful, not because the platform prevents it, but because the business has been optimized around remaining inside it.

Agentic commerce introduces the possibility that the same pattern moves upstream. Instead of owning the marketplace, the platform owns the conversation that determines what the customer sees before a marketplace is ever opened.

The strategic risk is not simply that a platform takes a fee. The deeper risk is that the platform becomes the place where the customer relationship is interpreted. The customer asks a question, the agent answers, and the brand appears only in the form the ecosystem allows. 

For brand-led retailers, this is the sharp edge of the model. Their value has never come solely from having products available. It comes from how those products are presented, explained, and experienced. When those interactions move into a platform-controlled environment, the brand may still be present. But presence is not the same as control.

Future Two: Rails and Rent

The second future is more subtle because the retailer may appear to retain more independence. The customer still recognizes the brand. The retailer may still own the customer record, operate the transaction, and manage the relationship after purchase.

Yet every interaction passes through infrastructure owned by someone else.

This is the rails-and-rent scenario. The defining characteristic is not loss of control. It is the emergence of a new economic layer. A platform creates value by helping customers discover, evaluate, or purchase products, then captures a portion of that value each time the interaction becomes a transaction.

The distinction matters. Ecosystem lock-in is a control problem. Rails and rent is an economic problem. One determines who shapes the relationship. The other determines who participates in the revenue generated by it.

At low volume, the cost may feel negligible. At scale, it becomes structural. A fee attached to a small number of transactions is a cost of experimentation. A fee attached to a meaningful share of demand becomes part of the retailer’s margin profile.

The clearest recent example appeared in September 2025, when OpenAI launched Instant Checkout inside ChatGPT. Merchants paid a fee on completed purchases, while transactions ran through the Agentic Commerce Protocol. For the first time, a major AI platform was not simply influencing purchase decisions. It was charging brands to complete transactions on its surface.

By early 2026, OpenAI had shifted course. Instant Checkout moved away from completing purchases inside ChatGPT and toward product discovery, with transactions returning to merchant systems. The reasons are revealing. OpenAI had not built the infrastructure required to manage sales tax collection and remittance across jurisdictions. Etsy reportedly agreed to absorb commission costs on behalf of merchants to encourage participation.

The significance of that experiment is not whether it succeeded or failed. It is that the model exists. An agent earning a fee for facilitating a purchase is no longer a hypothetical scenario. Future platforms may refine the approach, but the underlying idea is now on the table.

The concern is not a single fee. It is the possibility that investments in brand, product, and customer experience increasingly create value for someone else’s infrastructure. The stronger the relationship a brand builds with its customers, the more attractive that relationship becomes as a source of transaction revenue for the platforms sitting in between.

Future Three: The Retailer-Owned Model

The third future starts from a different premise. Instead of asking which external platform will control the agentic layer, it asks what the agent should be connected to in the first place.

In this model, the agent is not the destination. It is an interface. Its value comes from the systems, context, and relationships that already exist inside the retailer’s business.

This creates a different role for the agent. Rather than mediating the relationship between retailer and customer, it extends it. The agent becomes another way for customers to engage with the brand, much like a website, a mobile app, or a store associate.

A general-purpose shopping agent may understand a person’s behavior across thousands of purchases. A retailer-connected agent understands the history of a specific relationship. It knows what the customer bought previously, which products were returned, which store they prefer to visit, and which associate helped them last time. A general-purpose agent understands the customer as a shopper. A retailer-connected agent understands the customer as a customer.

NewStore CEO Mike DeSimone has written about the store as the one channel where a brand-led retailer still owns the relationship outright. The third path takes that idea one layer deeper, into the systems that make the store work.

Point of sale, customer history, order management, store operations, and clienteling platforms all contain pieces of context a generic agent cannot generate on its own. They are not new capabilities. They are existing capabilities exposed through a new interface.

The retailer-owned path does not mean building everything alone. Retail has never worked that way. Partnerships will remain necessary. Platforms will remain useful. Infrastructure will still matter.

What changes is the architecture. The value of an agent is shaped by what sits behind it. An agent can generate recommendations, compare products, and guide a conversation. The advantage does not come from owning the agent. It comes from owning what the agent depends on.

Where Brand-Led Retailers Draw the Line

These three futures will not arrive cleanly or separately. Most retailers will encounter elements of all three. A customer may discover products through a platform-controlled agent, complete a purchase through third-party rails, and ultimately receive service from systems the retailer owns.

For brand-led retailers, however, the distinction between these futures matters. Their value has never come solely from access to demand. It comes from the ability to shape experiences, build relationships, and operate a business according to their own model. The more those capabilities move into external platforms, the harder they become to differentiate.

This is what gives the third path its weight. It does not depend on rejecting agents, platforms, or new forms of commerce. It assumes they will become part of the landscape. The retailer’s operating model remains the foundation, while agents become another way for customers to interact with it.

NewStore has spent the last decade building the systems that generate that operational truth inside stores. When the third path emerges, it will not begin with a new interface. It will begin with the systems retailers already rely on to run the business.

Agentic commerce is often described as a technology transition. That is true, but incomplete. The more important transition may be structural. Technology determines what becomes possible. The structure determines who benefits when it does.