Agentic commerce: when the buyer is an AI agent.
Agentic commerce is commerce where AI agents carry out purchases on a buyer’s behalf — finding products, checking prices and availability, and completing orders through structured protocols. Your catalog got them to your door. This is what happens next.
Last updated:
What is agentic commerce?
For twenty years, B2B eCommerce assumed a human at the keyboard. Agentic commerce breaks that assumption: the buyer is software, acting under a human’s rules. An agent does not browse your homepage or admire your photography — it queries your catalog, retrieves the price its company negotiated, checks whether you can ship this week, and places the order through an API.
That changes what a “storefront” is. The sellers who win agent-driven orders are not the ones with the best-looking site — they are the ones whose pricing, availability, and ordering are machine-readable, authenticated, and governed by clear guardrails.
We prepare B2B manufacturers and distributors for exactly that — on Shopify Plus, connected to the ERP and PIM you already run.
Talk to usRelated: AI for B2B Commerce & Operations — making your catalog readable by AI. This page covers the next step: AI agents that buy.
How do AI agents actually buy?
Agent purchasing is not one behavior — it is a spectrum of delegation, and each stage raises the bar on what your systems must expose. Assisted buying only needs your catalog to be findable and accurate. Delegated buying needs authenticated access to your pricing and a way to place orders. Autonomous buying needs all of that plus guardrails a finance team will trust. Most B2B sellers can serve the first stage today and none of the others — which is where the work is.
Assisted buying
A human drives; the agent does the legwork. The buyer asks ChatGPT, Perplexity, or Claude to find and compare products, and approves every step before anything is purchased. This is where most buying activity sits today — and where your catalog either shows up with accurate pricing and availability, or does not show up at all.
Delegated buying
The human defines the task; the agent executes it. "Reorder our usual consumables when stock drops below two weeks of cover" is a delegated purchase. The agent needs authenticated access to contract pricing, live availability, and an order API — and the seller needs to recognize which company the agent represents.
Autonomous buying
The agent transacts on its own inside preset rules: approved suppliers, spending limits, budget periods. No human touches the individual order. This is the end state procurement teams are building toward — and it only works with sellers whose systems expose machine-readable pricing, availability, and ordering with proper guardrails.
What plumbing makes agentic commerce work?
Under every agent purchase sits a protocol — a structured way for software to discover, price, and buy. Five matter right now, and they arrived fast. You do not need to adopt all of them; you need to know which ones your buyers’ agents will use, and expose your catalog accordingly.
MCP — Model Context Protocol
The open standard that lets AI agents connect to external systems — catalogs, inventory, pricing, order APIs. Introduced by Anthropic in November 2024, it has become common plumbing for agent-to-system integration, which makes it the natural layer for exposing B2B commerce capabilities to agents.
Instant Checkout — Agentic Commerce Protocol
OpenAI launched Instant Checkout in ChatGPT in September 2025, letting buyers complete purchases without leaving the conversation. It is built on the Agentic Commerce Protocol developed with Stripe, rolled out first with Etsy sellers and then Shopify merchants.
Perplexity shopping
Perplexity launched its shopping experience in November 2024 — product discovery and purchasing inside an AI answer engine. An early signal that AI assistants would not stop at recommendations.
Shopify Agentic Storefronts
Launched March 24, 2026: Shopify merchants can sell directly inside AI assistants. D2C only — but hybrid B2B/D2C stores on Shopify Plus get AI visibility through their consumer catalog while B2B pricing stays private. This is the architecture we build for distributors.
UCP — Universal Commerce Protocol
Google’s open standard for agent-driven transactions across the web. Structured product feeds compatible with UCP are part of how we prepare catalogs for whichever agent surface a buyer happens to use.
What does agent-ready B2B look like?
B2B is where agentic commerce pays off first, because B2B buying is repetitive, specification-driven, and contract-governed — exactly what agents are good at. But it is also where the infrastructure gap is widest. Five capabilities separate sellers agents can buy from and sellers they route around. We have built these on Shopify Plus with ERP integration across 50+ B2B storefronts since 2018, and our B2B Protocol app — client-exclusive today, public Q3 2026 — extends B2B pricing to the point of sale.
Machine-readable pricing
Contract and tier pricing exposed through authenticated APIs — not locked in PDFs, spreadsheets, or a salesperson’s inbox. An agent that cannot retrieve the right price for the right company cannot buy.
Real-time availability
Live inventory and lead-time APIs. Agents deprioritize sellers whose stock status is stale or missing — a delegated reorder cannot wait for a human to “check with the warehouse.”
Punch-out-style flows for agents
B2B procurement has run punch-out catalogs for decades. Agents need the same idea rebuilt on modern APIs: authenticate, browse at negotiated prices, return a structured cart, place the order.
Agent authentication & guardrails
Know which company an agent represents, what it is allowed to buy, and how much it can spend. Guardrails are what make autonomous purchasing safe enough for finance teams to sign off on.
Clean order APIs
Order placement, status, and invoicing through APIs — not email threads. Our Husky Boutique build runs 40+ API endpoints with automated order routing to the ERP; that is the shape agents expect.
What must B2B sellers prepare?
Six steps, in order. Most sellers discover that steps one through four reuse integrations they already paid for — the ERP connection, the PIM, the storefront APIs. What is missing is the agent-facing surface on top of them.
Audit your machine experience
Walk your own purchase path as an agent would: can software retrieve catalog, pricing, and availability without a human? Every point where the answer is no is a lost agent transaction.
Structure your product data
Complete attributes, consistent identifiers, JSON-LD markup, and clean feeds. Discovery readiness comes first — an agent cannot buy what it cannot evaluate.
Expose pricing behind authentication
Publish contract and tier pricing in machine-readable form, gated by customer identity. Public list prices for visibility; private negotiated prices for authenticated agents.
Open availability and order APIs
Real-time stock, lead times, order placement, and order status as APIs. If your ERP already integrates with your storefront, most of this plumbing exists — it needs an agent-facing surface.
Define agent policies
Decide which agents you accept, how they authenticate, and what spending guardrails apply. Sellers who define this early set the terms; sellers who wait inherit whatever the platforms decide.
Pick your protocol bets
MCP, the Agentic Commerce Protocol, UCP, Shopify Agentic Storefronts — you do not need all of them on day one. You need a map of which ones your buyers’ agents will actually use.
Frequently asked questions
What is agentic commerce?
Agentic commerce is commerce where an AI agent — not a human clicking through a website — carries out part or all of a purchase on a buyer’s behalf. The agent finds products, compares options, checks price and availability, and completes checkout through structured protocols instead of web pages. It spans a spectrum: assisted buying, where a human approves each step; delegated buying, where the agent executes a defined task like reordering; and autonomous buying, where the agent transacts within preset rules and spending limits. The infrastructure arrived quickly: Anthropic introduced the Model Context Protocol in November 2024, Perplexity launched its shopping experience the same month, OpenAI shipped Instant Checkout in ChatGPT in September 2025, and Shopify launched Agentic Storefronts in March 2026. For sellers, the practical question is no longer whether agents will buy — it is whether your systems let them.
How is agentic commerce different from making my catalog readable by AI?
They are two halves of the same shift. Making your catalog readable by AI — structured product data, schema markup, Generative Engine Optimization — determines whether agents can find and evaluate you. Agentic commerce covers what happens next: whether the agent can actually transact. That means checkout protocols, real-time availability, machine-readable pricing, agent authentication, and clean order APIs. A catalog can be perfectly discoverable and still be a dead end if the purchase path requires a human filling out forms. We treat them as sequential workstreams: discovery readiness first, transaction readiness second. Our AI for B2B Commerce & Operations service covers the discovery half; this page covers the buying half. Most B2B sellers we audit have gaps on both sides — but the transaction gaps are usually cheaper to close, because they build on order APIs and ERP integrations that already exist.
Which platforms support agentic commerce today?
Several, each with a different scope. OpenAI’s Instant Checkout, launched in September 2025, lets buyers complete purchases inside ChatGPT; it is built on the Agentic Commerce Protocol developed with Stripe and rolled out first with Etsy sellers, then Shopify merchants. Perplexity launched its shopping experience in November 2024. Shopify launched Agentic Storefronts on March 24, 2026, exposing D2C catalogs to AI assistants — hybrid B2B/D2C stores get visibility while B2B pricing stays private. Google is developing the Universal Commerce Protocol (UCP) as an open standard for agent-driven transactions. Underneath several of these sits the Model Context Protocol (MCP), the open standard Anthropic introduced in November 2024 that lets AI agents connect to external systems like catalogs, inventory, and order APIs. We track each protocol and map which ones matter for a given seller’s channel mix.
Does agentic commerce work for B2B, or only D2C?
It works for B2B, but almost no B2B seller is ready — which is exactly the opportunity. Consumer-facing platforms moved first: Shopify Agentic Storefronts, for example, support D2C sales only, though hybrid B2B/D2C stores on Shopify Plus get AI visibility through their consumer catalog while contract pricing stays private. B2B buying is actually a better fit for agents than consumer shopping — it is repetitive, specification-driven, and governed by contracts rather than impulse. But it demands infrastructure most sellers lack: machine-readable contract and tier pricing, real-time availability APIs, punch-out-style flows an agent can traverse, authentication that recognizes which company an agent represents, and spending guardrails. We build this on Shopify Plus with ERP integration — the same architecture behind the 50+ B2B storefronts we have delivered since 2018, including Husky Boutique’s 40+ API endpoints and automated order routing.
How should a B2B seller start preparing for agentic commerce?
Start with an audit of how a machine experiences your business today. Can an agent retrieve your catalog, your customer-specific pricing, and live availability through an API — or does every step assume a human with a login and a browser? From there, the preparation list is concrete: publish structured product data, expose contract and tier pricing in machine-readable form behind authentication, provide real-time inventory and lead-time APIs, define agent authentication and spending guardrails, and make order placement and status available through clean APIs rather than email. Founded in 2018 in Montreal, we have shipped this pattern across 50+ B2B storefronts as a Shopify Plus Partner and Akeneo Partner, and our founder Rudy Abitbol spent 18+ years in B2B commerce, including as Director of Digital Channels at Lumen (part of Sonepar). The first engagement is a readiness audit, not a rebuild.