When Merchants Start Taking Back the Interface
Issue 19 examined the embedded commerce surface layer, where AI assistants, payment applications, search systems, and support agents are becoming places where customers can discover products, make decisions, and complete transactions. That shift moved commercial power away from the merchant’s website and toward whichever interface held the customer’s attention, context, and permission to act. Issue 20 examines the response now forming beneath that shift: merchants, commerce platforms, and regulators are beginning to define who may represent a business, which data machines can use, how AI interactions must be disclosed, and how much control a merchant must retain when commerce moves into external systems.
The central problem is no longer simply whether a merchant can participate in AI commerce. Participation without control can create a weaker business. A company may become visible across more interfaces while losing authority over product presentation, customer relationships, recommendation logic, data access, and the rules under which a transaction occurs. The same infrastructure that expands distribution can turn the merchant into a supplier of inventory and information beneath someone else’s commercial experience.
The developments between 28 July and 3 August show that the next layer of AI commerce will be built around merchant sovereignty. This does not mean forcing customers back to a conventional storefront. It means preserving enough control for businesses to decide how their products are represented, which systems can act on their behalf, what evidence accompanies an AI interaction, and whether the merchant can adapt without waiting for a dominant platform to update its rules. The next competitive advantage will not belong only to the business that reaches the most AI surfaces. It will belong to the business that can travel across those surfaces without surrendering the commercial system underneath.
Shopify Is Turning External Commerce Into an Accountable Partner Channel
On 28 July, Shopify explained how it is changing the incentives around its partner ecosystem. The company said partners will be able to use Shopify Catalog to build commerce experiences beyond traditional storefronts and earn commissions on the sales those experiences generate. Shopify also reported that its broader partner ecosystem generated $6.86 for every dollar Shopify earned in 2025 and supported more than 1.5 million jobs.
The important shift is not simply that developers can build more shopping applications. Shopify is creating a governed commercial layer between merchants and external interfaces. A developer may build an AI travel planner, gifting assistant, home-design tool, or specialist shopping agent that draws from Shopify’s product catalog. Instead of treating that application as an uncontrolled source of traffic, Shopify can identify the partner, connect the resulting transaction to merchant infrastructure, and compensate the builder for producing the sale.
This creates a model for preserving merchant participation without requiring the merchant to own every customer-facing interface. The merchant remains connected through the catalog, checkout, order, and operating systems underneath the experience. The external developer controls the use case and interaction, but does not need to reconstruct the merchant’s commercial machinery. Shopify, meanwhile, becomes the layer that determines how supply enters the experience and how economic value is divided.
The structural consequence is that commerce platforms are moving from software provision into channel governance. Their advantage will not come only from hosting stores. It will come from defining the technical and financial terms under which third parties may create demand from merchant inventory. The platform can establish permissions, attribution, commissions, transaction standards, and developer accountability across thousands of new AI-powered surfaces.
The second-order implication is that merchant sovereignty may increasingly be exercised through platforms rather than independently. A merchant gains reach and operational consistency, but the platform decides which external experiences qualify, what data they can access, how transactions are recorded, and how commissions are calculated. This can protect the merchant from fragmented integrations while strengthening the platform’s position as the unavoidable commercial intermediary.
Mental Model Update: External AI distribution stops being an open referral channel when a commerce platform can identify the builder, govern the transaction, and decide how the resulting value is shared.
Source: Partner Economy
Retailers Are Rebuilding the Decision Surface Inside Their Own Systems
Kohl’s published details of its new AI shopping assistant on 29 July. Built using Google Cloud’s Gemini Enterprise for Customer Experience, the assistant operates across Kohls.com and the Kohl’s app. It can provide personalised recommendations, search through uploaded images, compare products, surface offers, add items to the cart, answer policy questions, check store availability, support pickup, and help customers track orders.
The obvious interpretation is that Kohl’s has added conversational shopping. The deeper signal is that retailers are responding to external AI interfaces by building their own systems that combine discovery, comparison, conversion, and support. Rather than allowing a general assistant to interpret the catalog, choose which details matter, and send the customer only at checkout, Kohl’s is placing the conversational layer directly over its own product, promotion, availability, service, and order data.
That gives the retailer a different form of control. The assistant can understand which discounts are currently valid, whether a product is available at a preferred store, how pickup works, and where an existing order sits. A general shopping agent may compare more merchants, but it is unlikely to hold the same depth of retailer-specific operational context. Kohl’s is betting that a useful owned assistant can keep customers inside its environment by connecting inspiration with execution more accurately than an external interface can.
The broader change is that the storefront is evolving from a collection of pages into a retailer-controlled decision system. Product search, recommendations, promotions, customer service, and fulfilment information no longer need to remain separate functions. The assistant becomes a common interface through which the retailer can coordinate them around the customer’s current need.
The second-order implication is that merchants may operate two parallel AI strategies. The first makes products understandable and purchasable through external agents. The second creates an owned agent that uses richer internal data to deliver a more complete experience. Businesses that pursue only the first strategy may gain distribution but weaken their direct customer relationship. Businesses that pursue only the second may retain control but become invisible when customers begin elsewhere.
Mental Model Update: The merchant website does not regain power by adding another chatbot; it regains power when its AI can connect customer intent to inventory, offers, policy, fulfilment, and service inside one controlled environment.
Source: Kohl’s Assistant
Small Retailers Are Being Brought Online Through a Single Platform Stack
The Retailers Association of India and Google announced a collaboration on 30 July to support digital adoption among small and medium-sized retailers. The initiative covers RAI’s network of more than 600,000 storefronts and brings together Google Business Profile, Merchant Center, Shopping, Ads, Maps, Pay, and Cloud. The programme is intended to improve merchant discoverability, expose local inventory, support customer acquisition, and prepare retailers for AI-driven commerce.
This is structurally important because machine commerce cannot include merchants that remain digitally incomplete. A local retailer may have trusted products, loyal customers, competitive prices, and available stock, but an AI system cannot reliably select that business if its identity, location, catalog, inventory, and operating details are not represented in structured digital systems. Digitisation is therefore becoming a requirement for economic visibility, not merely a marketing upgrade.
Google can help close that gap because it controls several connected surfaces. A merchant profile can establish business identity and location. Merchant Center can carry structured product information. Shopping and Ads can create discovery. Maps can connect local intent with physical availability. Google Pay can participate in the transaction. Cloud and AI services can support the systems underneath. For small retailers, the advantage is that these capabilities can be adopted as one connected path rather than assembled independently.
The risk is that digital inclusion can become platform dependence. When a retailer’s identity, discovery, product visibility, advertising, local traffic, and payment access are concentrated within one ecosystem, the business may become technically visible while losing bargaining power over how that visibility is governed. Changes to data requirements, ranking rules, advertising economics, account policies, or AI recommendations can affect the merchant across several commercial functions at once.
The second-order implication is that the AI-commerce divide will not simply separate digital retailers from offline ones. It may separate merchants that own portable commercial data from those whose digital existence is largely defined inside one platform. The former can move their identity, inventory, and customer relationships between systems. The latter may become machine-readable only within the ecosystem that digitised them.
Mental Model Update: Digitisation stops being commercial independence when the same platform controls the merchant’s identity, product visibility, customer acquisition, transaction access, and path into AI commerce.
Source: Retail Digitisation
AI Transparency Has Become a Commerce-System Requirement
Article 50 of the European Union’s AI Act began applying on 2 August. It requires providers of AI systems that interact directly with people, including chatbots and AI agents, to ensure users are informed that they are dealing with an AI system unless that fact is obvious. Providers of generative AI systems must also apply robust and interoperable machine-readable markings to generated or manipulated outputs. The obligations apply to providers outside the EU when their system’s output is used within the EU.
For commerce, this turns disclosure from a user-interface preference into infrastructure. Retailers deploying shopping assistants, support agents, recommendation systems, synthetic product imagery, generated promotional content, or automated customer communication must understand where AI is operating and whether the required disclosure travels with the interaction or output. A label buried in general terms is not necessarily enough. The Commission says users should be informed clearly from the start of the first interaction when the obligation applies.
The machine-readable marking requirement is equally important. AI-generated content may move through marketplaces, advertising systems, merchant feeds, social platforms, and external shopping agents. A visible disclosure can be removed when an asset is reformatted or republished. A machine-readable mark is designed to allow downstream systems to detect that the content was generated or manipulated by AI. This introduces provenance into the commercial content supply chain.
That changes how businesses must manage product and marketing assets. A retailer can no longer treat AI-generated images, text, video, and conversations as ordinary content with a different production method. The business may need to preserve metadata, maintain records of how an asset was created, distinguish automated content from substantively reviewed content, and ensure disclosures remain intact as material moves across channels.
The second-order implication is that transparency can influence machine trust. Shopping agents, marketplaces, regulators, and consumers may eventually use provenance signals to evaluate claims, distinguish authentic product imagery from synthetic representations, and identify whether a recommendation came from a machine or a person. Compliance infrastructure may therefore become part of merchant reputation and product verification, not merely a legal cost.
Mental Model Update: AI disclosure stops being a sentence in a privacy policy when the identity of the machine and the origin of its content must remain detectable across the commercial system.
Source: Article 50
Platform Architecture Is Becoming a Merchant-Control Decision
WooCommerce surfaced an IDC-sponsored study examining how agentic commerce is changing platform requirements. The research argues that traditional commerce platforms were built for human browsing rather than machine buyers and identifies closed ecosystems, slow release cycles, restricted data access, and platform pricing as structural constraints. It reports that 65% of digital leaders identify legacy platform rigidity as a major barrier to scaling AI in commerce and frames open architecture as a way for merchants to adopt new models, payment methods, protocols, and integrations without waiting for a vendor’s roadmap.
The significance is not that open-source software is automatically superior. It is that the speed of AI-commerce change is exposing the strategic cost of restricted architecture. New agent protocols, identity systems, content rules, payment methods, and disclosure requirements are appearing faster than many platform release cycles. A merchant whose critical data and workflows cannot be accessed or modified may understand what needs to change but remain unable to implement it.
This transforms platform selection from a technology procurement decision into a question of commercial authority. Businesses must evaluate who controls their product data, customer records, checkout logic, AI integrations, and ability to move between channels. Convenience at launch can become constraint later if every new capability requires platform approval, a premium subscription, or migration to another part of the vendor’s ecosystem.
The second-order implication is that portability will become a measurable asset. Merchants will increasingly need catalogs, policies, customer permissions, transaction records, and AI instructions that can move between owned assistants, external agents, marketplaces, and payment environments. A platform that helps the merchant distribute these assets without losing governance will be more valuable than one that merely offers the largest number of built-in AI features.
Mental Model Update: A commerce platform stops being infrastructure you rent when its architecture determines which AI markets you may enter, how quickly you may adapt, and whether your commercial data can leave.
Source: Open Architecture
The System That Is Emerging
The week’s developments reveal a merchant sovereignty layer forming beneath AI commerce. Issue 19 showed that transactions are moving into interfaces merchants do not own. Issue 20 shows that the market will not accept this movement as a completely open transfer of control. Commerce platforms are building governed partner channels. Retailers are deploying their own decision systems. industry associations are digitising merchants through integrated ecosystems. Regulators are making machine identity and content provenance explicit. Open-commerce providers are turning architecture and portability into strategic issues.
The emerging layer determines which commercial rights survive when the storefront is no longer the only place where commerce happens. Those rights include:
- the right to define and update the authoritative product record;
- the right to know which machine or partner is representing the business;
- the right to control access to inventory, pricing, policies, and customer data;
- the right to preserve attribution and economic participation when others create the interface;
- the right to identify AI-generated interactions and commercial content;
- the right to move data and business logic when a platform no longer serves the merchant.
The old model assumed these controls existed naturally because the merchant owned the website and checkout. Embedded commerce separates them. One company may control the interface, another the product catalog, another the customer identity, another the payment, and another the fulfilment. Sovereignty must therefore be designed into the connections between systems rather than assumed from ownership of a storefront.
Control is moving toward platforms that can balance reach with governance. Merchants want access to AI-driven demand, but they also need an authoritative commercial core that external experiences cannot silently overwrite. Developers want freedom to build new buying journeys, but merchants need identity, permissions, attribution, and enforceable terms. Regulators want innovation, but they also require people and machines to know when AI is acting and when content has been generated.
Core Truth: The merchant that becomes available everywhere but authoritative nowhere has not gained distribution; it has surrendered the commercial system that makes distribution valuable.
For operators, this creates a practical test. Every AI-commerce initiative should be assessed not only by the demand it might create, but by the control it changes. Who owns the customer interaction? Which system holds the authoritative product and policy data? Can the merchant identify the agent or partner? Does attribution survive? Can the business retrieve its records, change providers, or revoke access? Does AI-generated content remain traceable? The answers will determine whether an AI channel expands the business or quietly makes it more dependent.
