AI agents add a distribution task for brands

A customer can prefer a brand and still struggle to buy it through an AI agent if the system cannot understand its products or complete the purchase. Agentic commerce, shopping in which AI systems act on a consumer’s behalf, adds a distribution task: making products usable in the AI environments customers choose. Incomplete, inconsistent, or inaccessible product information can block a purchase even when marketing has persuaded the customer. Brands therefore need to connect customer preference with systems that can act on it.

Supporting those systems gives marketing, commerce, and engineering teams a shared problem. Marketing creates demand, while commerce and engineering make the offer available and the purchase possible through the customer’s chosen interface. Websites, apps, and social channels continue to serve those functions as AI environments develop alongside them. To assess the extra work, teams first need to understand how an agent changes the customer journey.

Conversational access changes where customers act

The conventional online journey brings customers into an environment the brand controls. A brand attracts visitors to its website, app, or online store, then combines visual storytelling, emotional copy, product information, and purchasing functions to encourage action. Click-through behavior matters because reaching that environment is an important step toward conversion. The storefront brings the presentation and the next action together.

An agent-mediated journey moves that interaction into an environment the consumer chooses. The customer describes an intention in conversation, and the assistant helps discover products or services, compare options, and carry out supported actions or purchases. The brand’s offer must therefore be accessible within the assistant’s environment. For a supported task, the customer can reach the service without visiting the brand’s site or app.

Spotify illustrates this change in access to a digital service. Through Muse, which Meta describes as its latest personal agent, users can make conversational requests and have the agent interact with Spotify for them. Both companies have a commercial interest: Spotify benefits when people use its service, while Meta benefits when they use its agent as their interface. Spotify’s announcement states: “Your agent can play music on Spotify, save songs to your Library, build playlists, help you find a book or podcast, and more – all from a conversation.”

These functions let users request music, a playlist, a book, or a podcast without manually opening Spotify and navigating its interface. Spotify still supplies the service, while another company provides the conversational entry point. Retail commerce applies a similar pattern to finding products, comparing offers, and completing purchases. To complete that task, the assistant needs connections to product information and checkout.

Okoone experts
LET'S TALK!

A project in mind?
Schedule a 30-minute meeting with us.

Senior experts helping you move faster across product, engineering, cloud & AI.

Please enter a valid business email address.

Catalog access and shared standards connect agents to merchants

Those connections become easier to build when merchants and AI environments share a framework for commerce interactions. In early 2026, Shopify announced the Universal Commerce Protocol, or UCP, calling it a “new open standard co-developed with Google to bring commerce to agents at scale.” Shopify supplies commerce infrastructure and benefits when merchants use its services to reach additional sales channels. Its description of UCP reflects that commercial interest in expanding commerce through AI environments.

Google also has an interest in making its AI services useful places to shop. Ashish Gupta, VP/GM of Merchant Shopping at Google, said the shift would “require a shared language across the ecosystem and the Universal Commerce Protocol provides that framework.” A shared framework gives merchants and AI environments a common way to connect shopping requests with supported commerce functions. Where those functions are available, the conversation can lead to a transaction within the assistant’s environment.

To participate, merchants also need a route for their products into those environments. Shopify says its agentic plan lets brands on other commerce platforms make products available through Shopify Catalog, which supported AI channels can then offer for sale. Catalog distribution is therefore separate from the platform running a brand’s storefront. A brand can assess access to those channels while keeping its existing commerce platform and direct route to customers.

Catalog access creates an opportunity only where a channel supports the actions customers want to take. Shopify describes its rollout as including “native shopping” in Google’s AI Mode and Gemini, checkout embedded in Microsoft Copilot, and integrations with ChatGPT. These arrangements give participating merchants more places to reach consumers, but their functions and rollout stages differ. When assessing a channel, a brand therefore needs to distinguish product discovery, comparison, and supported purchasing within that experience.

Checkout connects product discovery to a purchase

The distinction between those functions becomes clear when a customer shops across several retailers. Google describes Universal Cart as spanning Search and Gemini and allowing consumers to add items from multiple retailers to a cart. Gemini monitors prices and availability and recommends alternative products. Comparison and purchase preparation can therefore stay within the environment where the customer is already shopping.

Payment extends that activity into checkout for participating retailers. Through UCP, those retailers can support direct payment from the cart using Google Pay. Nike, Sephora, and Walmart are set to support selected checkout functions. Their announced participation is forthcoming and applies to particular functions, so the purchasing experience depends on both the retailer and the supported service.

Google’s existing shopping activity explains the commercial appeal of keeping more of this journey within its services. The company says people use its platform to shop “more than a billion times a day.” That figure measures shopping activity across Google. Google describes its work as “building the foundation for agentic commerce,” with UCP providing a shared framework for the interactions involved.

For a brand, the practical question is which parts of a purchase each environment can carry out. Catalog access can make an offer available for discovery, while checkout support determines whether the customer can complete the transaction there. The Spotify example shows how a conversational interface can carry out supported service tasks; retail purchasing adds requirements for product selection, availability, pricing, and payment. Each added function places demands on the information and systems behind the offer.

Product data makes an offer usable by agents

Those demands make product information part of distribution strategy. A shared protocol gives systems a way to interact, but the brand still has to supply an offer they can understand and use. A product page designed for a human visitor may leave software unable to interpret important information if the underlying data is incomplete or hard to access. Preparation therefore extends behind the storefront into product records and the connections that make them accessible.

Agentic Commerce Optimisation, or ACO, describes this preparation for discovery and purchasing through agents. SEO, or search engine optimization, provides an established approach to visibility in search; ACO extends that work into product evaluation and supported transactions. The aim is for an agent to understand an offer, select it appropriately, and complete the requested purchase. That requires usable information at each step.

Usable information starts with structured metadata: product details organized in a form software can read. An agent needs to interpret those details against the customer’s request, and availability and pricing can influence whether it can fulfill that request. Inconsistent values or inaccessible information can therefore disadvantage a product even when its presentation works well for human visitors. Clear product records help connect what the customer wants with what the brand can supply.

Those records must also reach the agent in time to support a decision. An application programming interface, or API, lets software systems exchange information and request supported functions; its real-time performance affects access to current product information during shopping. A large language model, or LLM, is an AI system that processes and generates language; integration into LLM pipelines means making product information usable in the processing flows behind conversational systems. Together, the data and its delivery determine what the agent can evaluate when the customer asks it to act.

This dependence on brand-level information explains the separate roles of protocols and implementation. UCP provides a common framework across participating systems, while metadata, accessible product records, and responsive APIs make an individual offer usable through it. Incomplete information remains a problem even when systems share a protocol. Well-organized information, in turn, needs a supported connection through which an AI environment can discover the offer and perform the relevant commerce functions.

Teams need to connect preparation with purchasing outcomes

Making those connections work requires coordination across the teams that shape demand and fulfill it. Branding influences customers’ interests and the preferences they express in conversation, while commerce and engineering determine whether an agent can act on those preferences. Marketing teams therefore need to understand how products appear in supported AI environments, and technical teams need to understand the purchase requests those environments must fulfill. Product presentation and transaction readiness serve different stages of the same customer task.

Coordinating that work also means choosing a scope that fits each channel’s current capabilities. The shift toward agentic commerce is accelerating, but forthcoming and selective checkout support makes preparation dependent on the retailer and experience involved. Teams can assess whether a chosen environment can discover their products, interpret the relevant information, and carry out the intended action. This work can extend existing distribution, including through Shopify’s cross-platform catalog route, while storefronts continue serving customers directly.

Preparation for human and AI audiences is also expected to feature in expert discussions at DMWF’s upcoming conference in Amsterdam. For teams turning those discussions into implementation decisions, a specific customer task offers a useful starting point. They can examine whether an agent can find an appropriate offer, obtain current price and availability information, and complete the supported purchase. Each step gives the responsible team a concrete part of the experience to assess.

Assessing those steps calls for measures tied to what happens inside the AI environment. Possible measures include how often an agent selects a brand’s product and how often it successfully completes an automated transaction. These are proposed measures. They are not established industry standards. Selection would show whether an offer reaches the agent’s decision, while successful completion would show whether the purchasing path works. Tracking both would help teams distinguish access to a channel from a usable purchasing experience.

Main highlights

  • Distribution now includes AI environments: Brands need to make products accessible and usable through the assistants customers choose. Marketing, commerce, and engineering share responsibility for connecting customer preference with a supported purchase.
  • Conversational access moves the customer interaction: AI assistants can become the entry point for discovery, comparison, and supported actions. Commerce teams need to assess which customer tasks each environment can complete.
  • Shared standards expand channel access: Protocols such as UCP connect participating merchants and AI environments. Commerce leaders can assess catalog distribution options, including Shopify’s cross-platform route, while retaining their existing storefronts.
  • Checkout support determines transaction reach: Discovery, comparison, and payment capabilities vary by channel and retailer. Commerce teams need to verify supported checkout functions and rollout status before committing to a channel.
  • Product data makes offers usable: Structured metadata, accessible product records, and responsive APIs help agents evaluate offers and obtain current prices and availability. Engineering and commerce teams need to assess data quality alongside channel integration.
  • Measurement needs to follow the purchase: Marketing, commerce, and engineering teams can test a specific customer task from discovery through checkout. Product selection and successful transaction completion offer potential measures for locating gaps in that experience.

Alexander Procter

November 17, 2026

10 Min

Okoone experts
LET'S TALK!

A project in mind?
Schedule a 30-minute meeting with us.

Senior experts helping you move faster across product, engineering, cloud & AI.

Please enter a valid business email address.