Why B2B agentic commerce can’t run on retail protocols

What happened when retail protocols met real buyers?

Before determining if retail protocols are suitable for B2B, it’s important to clarify their intended purpose. Their specifications provide the answer.

ACP and UCP standardize consumer discovery and checkout

OpenAI’s Agentic Commerce Protocol is three specifications: a product feed, an agentic checkout flow, and a delegated payment spec. The feed’s required fields exist to display price and availability correctly, and the production guide states the protocol currently models a single shipping address. One buyer, one address, one posted price.

Google’s Universal Commerce Protocol covers the consumer shopping journey on Google surfaces. Its core capabilities are catalog search, cart building, identity linking, checkout, and order management, with customers paying through Google Pay using credentials stored in their Google Wallet.

Both protocols define the same transaction: a person finds a product at a listed price and buys it. If you’re evaluating which protocol powers AI shopping agents, that definition is the starting point, and it matters just as much when choosing between ACP and UCP for consumer channels.

Consumer adoption stayed low even where the protocols worked

The retail model behind these specs is still unproven. Forrester reports that US consumer adoption of Instant Checkout stayed low and stagnant from its September 2025 debut to its March 2026 shutdown, and that consumer interest in AI agents making purchases remains lukewarm.

The conversion data explains the decision. Walmart told Wired that purchases completed inside the ChatGPT chatbot converted at one-third the rate of purchases where users clicked out to its own site. 

None of this means agentic commerce is failing. It means fully automated consumer checkout has not yet proven demand. And that use case was never designed for B2B in the first place.

Why can’t a consumer protocol carry an institutional purchase?

McKinsey describes the difference plainly. In consumer commerce, a shopper authorizes an agent to save time or money. In B2B, authority comes from procurement policies, budget owners, risk teams, and legal frameworks. An agent acting in a B2B buying environment can trigger approvals, inventory commitments, and contractual obligations in a single action.

ACP and UCP model none of that institutional structure. They define one user with one payment credential authorizing one transaction. The mismatch runs through every part of the purchase:

The left column does not criticize the protocols; it simply reflects their design as confirmed by their own specifications. The real issue lies in the right column, where B2B buyer agents currently operate. Forrester reports that 61% of purchase influencers say their organizations have implemented or plan to use a private generative AI engine to aid in purchasing. Additionally, procurement teams are developing agents to negotiate discounts, payment terms, and service levels, all while ensuring compliance with regulations and entitlements.

Forrester anticipates that static pricing pages and inflexible checkout processes will be replaced by negotiation interfaces compatible with agents. A negotiation interface differs from a checkout interface. Retail protocols standardize the latter but don’t define the former.

Where is B2B agentic commerce actually being built?

Picture a procurement team that sends quote requests to a hundred suppliers overnight, scores every response against contract terms, and returns counteroffers before your sales team logs in. That is the workload Forrester’s negotiation prediction describes, and no retail checkout protocol is involved.

The research points to three signals that B2B agentic commerce is developing on its own routes:

1. Buyers are automating faster than their suppliers. 

Deloitte’s 2026 B2B commerce research, surveying more than 1,000 US suppliers and buyers, found that nearly 40% of B2B buyers already use agentic AI in purchasing: evaluating products, configuring orders, reviewing contracts, and benchmarking prices. Only 24% of suppliers use agents in their sales process. Your buyers aren’t waiting for a protocol standard, and they aren’t waiting for you.

2. The system being replaced is EDI, not the shopping cart. 

In retail, agentic commerce arrived as a new checkout button. In B2B, it is an infrastructure succession: Deloitte reports that 92% of buyers currently using EDI, the document-exchange standard that has carried B2B transactions for more than 40 years, plan to shift partially or fully to other channels. The migration points toward API-ready systems agents can connect to directly, one of the defining AI trends reshaping B2B e-commerce.

3. Agent payment infrastructure is arriving in B2B first. 

Forrester’s reasoning is specific: B2B payments suit agentic AI because the complexity is in adjacent processes, invoicing, accounts payable and receivable, and trade credit, and because these processes carry none of the consumer-trust and authentication problems the retail protocols exist to solve. Forrester also expects Stripe and Tempo’s new Machine Payments Protocol, built for agent-to-agent payments, to take hold in B2B first, where payments are highly repeatable.

So which standards apply here? Deloitte’s answer: for B2B, the protocol landscape is still forming, protocols are in early stages, and no clear leader has emerged. The exception is MCP, which standardizes how agents connect to systems rather than how a purchase works. Connectivity transfers between retail and B2B. Commercial semantics–the pricing, terms, and approval logic that separate a checkout from a procurement package–do not. That is the layer B2B is still standardizing.

How do B2B sellers prepare for agentic commerce?

Start with what agents can read. McKinsey‘s warning to retailers applies with more force in B2B: if your catalog, policies, and value proposition are not machine-readable, agents will not find you, no matter how strong your brand. 

And once agents can read the data, McKinsey expects B2B competition to move from unit price to predictability, transparency, and policy alignment. In most organizations, the product data already exists; AI just can’t read it.

Governance is the other half, and skipping it now has a cost. Forrester predicts ungoverned genAI in commercial applications will cost B2B companies more than $10 billion in enterprise value through declining stock prices, legal settlements, and fines, driven by untested functionality and lagging AI user skills across teams that already use genAI daily.

Buyers themselves point to the working model. Forrester expects them to turn to human experts to validate genAI insights and confirm an offering meets their requirements, tasks genAI cannot take on alone. Agents accelerate the work; validation determines whether the output can be trusted. That standard applies to your own stack too: if your PIM isn’t AI-ready, neither is your agentic commerce strategy.

Get your product data ready for B2B agentic commerce

The protocol race will settle without your vote. What you control is whether your product and commercial data are ready for the transaction B2B actually runs on: quotes, terms, and approvals, not carts. 

Inriver’s flexible data model ingests product data from your existing sources without an upfront restructuring or ETL project, so it becomes usable and AI-ready fast. Its agentic orchestration then runs AI agent and LLM workflows across product operations with verification and validation built in, so agent output is checked against trusted product data before it reaches a buyer or a channel. That is agentic speed without giving up control.

See how Inriver makes product data ready for B2B agentic commerce; schedule a personalized demo today.