S hopify made agentic storefronts generally available and now serves AI channels through Shopify Catalog by default.
OpenAI and Stripe’s Agentic Commerce Protocol has been in the market since September 2025. So the infrastructure is real. Here’s the part most articles skip: almost none of it handles a B2B transaction yet.
If you sell engineered products, replacement parts, or anything with account-specific pricing, this matters more than the headlines suggest. The gap between “agents can buy things” and “agents can buy things from you” is where your next two budget cycles get decided.
What is Agentic Commerce in B2B terms?
Agentic commerce is when software acts on a buyer’s behalf across the purchase journey: gathering requirements, comparing options, building a cart, and in some cases completing the order.
In B2B, that journey carries commercial rules a consumer purchase doesn’t. Contract pricing tied to an account. Entitlements that govern who can buy what. Approval chains. Purchase orders. Net terms. Punchout connections into a buyer’s procurement system.
Those rules are the whole transaction. Strip them out and you don’t have a B2B sale, you have a retail one with a company name on the invoice.
What Actually Shipped in 2026
Four standards now define most agent-to-merchant activity, and they layer rather than compete.
- UCP (Universal Commerce Protocol). Google and Shopify, launched January 11, 2026. Covers the full journey from discovery through post-purchase. Transport-agnostic: it runs over REST, GraphQL, JSON-RPC, MCP, and A2A. Merchants declare what they support and agents negotiate from there.
- ACP (Agentic Commerce Protocol). OpenAI and Stripe, September 2025. Narrower, focused on the checkout and payment step.
- AP2 (Agent Payments Protocol). Handles the cryptographic proof that a user authorized a purchase. UCP composes with it rather than replacing it.
- MCP (Model Context Protocol). The tool-access layer. It’s how an agent calls your systems at all.
Most merchants who care about distribution will end up supporting more than one. Large retailers already run ACP for ChatGPT and UCP for Gemini and Google AI Mode.
Why Agentic Storefronts Don’t Sell B2B Yet
This is the single most useful fact in this article, and it comes straight from Shopify’s documentation: agentic storefronts support direct-to-consumer sales only. Products sold exclusively to business or wholesale customers are automatically excluded from AI channels when Shopify can identify them.
For hybrid sellers, Shopify is blunt about the consequences. If you sell the same product to both B2B and D2C customers, the product shows up in your agentic storefronts at the D2C price. In Shopify’s words, “B2B pricing doesn’t display on agentic storefronts.”
There’s a related risk worth an hour of someone’s time this week. Shopify detects B2B-only products through B2B catalogs, customer account requirements, and password-protected storefronts. If your team manages wholesale pricing through a third-party app or a theme customization instead (hiding prices until login, for example), Shopify may not recognize those products as B2B, and they can surface on an agentic storefront unintentionally.
BigCommerce is in a similar place. Its MCP server is in beta, with a B2B storefront listed as coming.
The protocols themselves carry the same gap. Across the current specs, there’s no standard representation of customer-specific contract pricing, account hierarchy, entitlements, approval authority, purchase orders, or net terms. Nobody has published a machine-verifiable credential that proves an agent has authority to commit a company to a purchase. Until that exists, an external agent transacting autonomously against your contract pricing on credit terms isn’t a thing you can buy, from anyone.
Where B2B Agents Are Already Producing Results
The working deployments share a trait: the agent operates inside one organization’s systems and controls, where identity and commercial rules are already resolved.
Order intake. Manufacturers and distributors receive a large share of orders as emails, PDFs, and spreadsheets. Agents that parse those documents, match SKUs and quantities to catalog data, and route them to the right account and pricing structure are in production today. This is measurable against labor hours, cycle time, and entry errors, which makes the business case defensible.
Assisted quoting inside your own channel. Salesforce made Agentforce Commerce generally available in late June 2026, including a B2B buyer agent that supports reordering against current contract pricing over messaging channels. The reason it works is architectural: the buyer is authenticated and the commercial logic stays inside the platform.
Merchant and sales operations. Shopify’s Sidekick, BigCommerce’s Companion assistant, Adobe’s CX Enterprise Coworker, and Optimizely’s Opal all put agents next to the people running the store. Optimizely has gone furthest on the B2B side, with MCP work covering catalog search through order placement for its B2B commerce engine, plus a remote MCP server released in April 2026 that exposes experimentation and feature management to AI tools.
Procurement platforms. This is the channel most manufacturers underweight. Your buyers’ agents are more likely to reach you through Coupa, Ariba, or JAGGAER than through a chat window. SAP shipped Joule agents across Ariba and Fieldglass in June 2026. If your punchout catalogs and pricing feeds are stale, that’s where you lose the order, and no storefront work will fix it.
What the Forecasts Actually Say
Two numbers dominate every agentic commerce deck. Both deserve context.
Gartner, October 2025: by 2028, 90% of B2B buying will be AI-agent intermediated, moving more than $15 trillion of B2B spend through agent exchanges. Read “intermediated” carefully. It plausibly covers a buyer using a chatbot during research or a procurement tool re-ranking a vendor list. It is not a forecast that $15 trillion will be executed autonomously.
Gartner, March 2026: 67% of B2B buyers say they prefer a rep-free experience, up from 61% the year before, and 45% used AI during a recent purchase. That’s from a survey of 646 buyers.
But Gartner’s May 2026 research adds the counterweight: 69% of buyers go back to a sales rep to validate what AI told them, and buyers consulted an average of seven information sources on a recent purchase. Forrester’s mid-2026 assessment lands in the same place, describing B2B agentic use cases as real but early, with market narratives running ahead of actual behavior.
The honest read: AI is already shaping how buyers research and shortlist you. Autonomous agent purchasing in B2B is not yet the thing to plan your roadmap around.
A Readiness Checklist that Survives Whichever Protocol Wins
Every item below pays off regardless of which AI surface matters in 2028. That’s the point of the list.
- Pricing as a service, not a nightly batch. An agent can’t use contract pricing that resolves once a day at 2 a.m. If your ERP integration was built for batch sync, this is the work.
- Entitlements and account hierarchy exposed through an API. Who can buy what, at what price, under whose authority.
- Product data an agent can parse. Normalized attributes, fitment and compatibility rules, units of measure, stable identifiers, specs and documents attached to the right SKU.
- Inventory and lead times with honest labels. Distinguish what’s available from what’s estimated. Agents will treat both as fact.
- Punchout and procurement network coverage. Audit which customers still work around a missing connection.
- Your agent discovery files. Every Shopify store already serves
/agents.md,/llms.txt, and/llms-full.txt, and you can override them with theme templates. Most teams have never looked at what theirs say. Look. - Channel audit for hybrid B2B and D2C. Confirm no wholesale-only product is reachable through an agentic storefront at a D2C price.
- Governance. Name the owner of data quality, the approval thresholds, and the points where a human still signs off.
Platform Notes for Manufacturers
There’s no single right platform. The right answer depends on your workflows, your team’s operating model, and what your ERP can realistically expose.
Shopify and Shopify Plus give you the strongest consumer-side agentic distribution available right now, plus Catalog, Cart, and Checkout MCPs from the Spring ’26 Edition. For a hybrid manufacturer with a real D2C line, that’s genuine reach. For wholesale-only revenue, treat agentic storefronts as out of scope today and put the effort into data, entitlements, and procurement connections instead. Shopify Plus merchants running a separate expansion store for B2B have the cleanest separation.
BigCommerce is building along the same path with its MCP server in beta and B2B support still ahead of it.
Optimizely is the most B2B-native of the three for agent work, given how much of the commercial logic (contract pricing, entitlements, approval workflows) already lives in the platform rather than in bolt-on apps. Its MCP surface is aimed squarely at the B2B journey.
The differentiator isn’t the logo. It’s whether your pricing, entitlement, and inventory services can answer a question in under a second, from any channel.
A 90-day Plan That Doesn’t Require Guessing
- Weeks 1 to 2. Audit what machines see today. Pull your agent discovery files. Check whether any B2B-only product is exposed to AI channels. Review your punchout catalog freshness with your three largest customers.
- Weeks 3 to 6. Pick one internal journey. Order intake from emailed POs is usually the highest-value starting point, because you can measure it against labor and error rates rather than an assumed conversion lift.
- Weeks 7 to 12. Fix one data path properly. Move contract pricing or inventory from batch to API. Deloitte’s research found 87% of B2B suppliers are upgrading or planning an ERP upgrade, so this work may already have a budget line you can attach to.
Then measure. Time to quote, reorder rate, order exceptions, returns from wrong-part orders, manual entry hours, support escalations. Those metrics justify the next phase whether or not a third-party agent ever checks out on your site.
The Takeaway
Agentic commerce in B2B is real, unevenly. Seller-side and internal agents are producing measurable results now. Buyer agents work inside controlled, authenticated channels. Consumer-style agentic storefronts are live and largely closed to B2B pricing by design.
That’s not a reason to wait. It’s a reason to spend the next two quarters on the layer every version of this future depends on: pricing, entitlements, and product data that a machine can read and trust. That investment pays whether the order arrives through Gemini, through Coupa, or through the same rep who’s been handling the account for nine years.
