HAS the opening of the 2026 edition of NRF Retail’s Big Showannual meeting where global distributors, e-commerce platforms and payment players meet, Google presented the Universal Commerce Protocol (UCP), an initiative supported by Walmart, Shopify, Target, Wayfair, Etsy and payment networks and providers. This new protocol is intended to be a collective response to the irruption of agents into the act of purchasing. The rise of agents capable of searching, comparing and executing a transaction on behalf of a user raises a central question: how to transform an intention expressed in natural language into a real, reliable and compliant transaction, without multiplying specific integrations between each interface, each merchant and each payment provider?
By offering a common language between agents and merchant systems, the UCP protocol seeks to streamline the chain from discovery to purchase. However, a question remains: can a single standard really absorb the diversity of business models, local rules and regulatory constraints?
When the agent decides for the user
Online commerce is built around the navigation process of Internet users, with a visual comparison, an emotional arbitration, then the checkout. With agentic commerce, this customer journey is radically modified, all the crossing points could disappear, erased by an agent who will interpret the user’s intention, then compare the possible purchase options, to finally decide.
A change raises many questions, who defines the agent’s decision criteria, and to what extent will these criteria remain transparent for both the user and the merchant?
If the agent arbitrates, according to what priorities does he do so: price, deadline, margin, loyalty, durability, implicit commercial agreements?
Why checkout is no longer a simple “pay” button
In classic online commerce, checkout is an end point. Once the “pay” button is activated, the transaction is frozen: the products, the price, the delivery method and the payment method can no longer change without starting the purchasing process again.
The UCP introduces a different logic by making the checkout a dynamic object, which the agent can manipulate. Thus an agent can initiate an order with a product, check availability, apply a discount linked to a loyalty program, change the delivery method to respect a deadline constraint, then offer a payment method compatible with the user’s preferences, all without recreating the transaction.
Standardize the dialogue between agents and merchants, without standardizing commerce
Technically, the UCP is based on a standardized format, where the merchant indicates what he is able to offer: types of products, payment options, discount rules, delivery terms.
Today, each new interface (search engine, conversational assistant, application) must develop its own connections with each merchant to access catalogs, prices or payment terms. This model, based on multiple and specific integrations, mechanically slows down the rise of agentic commerce.
By proposing a common language between surfaces and merchants, the UCP seeks to simplify these exchanges: an offer described once can be understood and exploited by several agents, in different environments. Commerce is becoming more fluid, however at the cost of greater abstraction.
This simplification is not without raising new questions. When offers are described using common schemas, it is these schemas that structure the comparison. What is not formalized, explained or translatable into this common language risks having less weight in the arbitration of agents.
In other words, in a commerce driven by systems, the question is no longer only what a brand offers, but how this proposition is made readable and comparable by agents who reason based on standardized criteria.
The risk is not so much the standardization of products as that of their representation. If a commercial advantage (premium service, environmental commitment, return flexibility) is not formalized in a readable manner by an agent, it becomes de facto invisible. Conversely, what is well structured, explained and exposed gains weight in algorithmic arbitration.
In an agentic world, the battle will not be fought only on the offer itself, but on the way in which it is described and encoded.
Brands: seducing a customer who feels nothing
When the agent becomes the main contact, the traditional levers of visual and emotional marketing lose their immediate influence. The agent perceives neither the staging of a product page, nor the graphic identity of a brand, he reasons based on explicit criteria: price, availability, deadlines, conditions, reliability.
Let’s take a concrete case. Two brands offer an equivalent product. One historically focuses on a premium image and a polished experience. The other sets out its deadlines, guarantees and return rules very clearly. For an agent responsible for optimizing a purchase according to specific constraints, the second may appear more “readable”, regardless of notoriety or storytelling.
This leads to a strategic question: does agentic commerce promote a return to competition centered on price and performance, or does it on the contrary open the way to new forms of differentiation, encoded in rules, data and capabilities?
In other words, does the brand become a simple parameter among others in the agent’s arbitration, or does it retain a structuring role provided that its promises are translated into a language usable by autonomous systems?
Payment, consent and accountability in the age of agents
The UCP introduces a clear separation between the payment instrument used by the consumer (card, wallet, bank account) and the service provider responsible for processing the transaction. This architecture aims to preserve interoperability and prevent a single actor from capturing the entire payment chain.
In an agentic journey, the payment often becomes invisible, the user authorizes the agent to act within a given framework, and the agent executes the transaction. The question is therefore no longer just technical, but legal and cultural: how to ensure that consent remains explicit and understood when the act of payment is no longer materialized by a click or a dedicated interface?
If the promise of consent is verifiable (traceable and revocable authorizations), it does not yet resolve all the regulatory implications, particularly in terms of consumer protection and liability in the event of a dispute.
An open standard, but what governance?
Led by Google and supported by major players in commerce and payment, the UCP presents itself as an open and agnostic standard. Its first concrete implementation, however, is part of Google’s conversational environments, which raises a question of governance.
Who will decide on changes to the protocol? How can we arbitrate between the sometimes divergent interests of platforms, merchants and payment providers? The history of protocols shows that formal openness excludes neither power asymmetries nor the effects of progressive domination.
The neutrality of the UCP will be played out as much in its governance mechanisms as in the actual diversity of its implementations.
UCP: 10 questions to ask yourself
-
Who will control access to the customer tomorrow when the purchase will be orchestrated by agents rather than interfaces?
-
What elements of our value proposition will remain visible and differentiating in a standardized business read by agents?
-
What level of autonomy are we prepared to grant to agents in the purchasing decision without diluting our legal and commercial responsibility?
-
Is our organization ready to move from interface commerce to commerce based on capabilities, rules and data?
-
What is the strategic cost of not participating now in the structuring of these standards?
-
How can we expose our commercial rules (prices, discounts, logistics) without losing control of their interpretation by third parties?
-
What signals will agents actually use to decide between comparable offers, and are we visible on these criteria?
-
Are we able to trace, explain and challenge an automated purchasing decision in the event of a dispute or non-compliance?
-
How much of our value still lies in the interface, and how much needs to migrate to the underlying infrastructure and rules?
-
Can we build a sustainable competitive advantage by becoming agent-ready earlier than our competitors?
