Agentic Commerce

Agentic commerce is what happens when AI stops recommending products and starts buying them. In this model, an AI agent receives a goal from a human buyer, then autonomously handles the entire purchase: researching options, comparing prices, negotiating terms, selecting fulfillment, processing payment, and completing the order. The human reviews and authorizes. The machine does everything else.

This is not a chatbot answering product questions. Forrester's working definition pins the distinction: agentic means the system flexibly plans, adapts, and takes action to resolve a goal with increasing autonomy. In commerce, that means the agent is the buyer. The human is the approver.

Where Agentic Commerce Stands Right Now

Forrester is blunt about the gap between narrative and reality: "Agentic commerce only sort of exists." Most "agentic" shopping experiences in mid-2026 are still conversational. A user asks ChatGPT about wireless headphones, gets recommendations, and finishes the purchase somewhere else. True autonomy, where the agent handles checkout without direct oversight, remains rare in consumer retail and early-stage in B2B.

But the infrastructure is moving fast. OpenAI charges a 4% transaction fee on ChatGPT Shopping purchases. Shopify has made 5.6 million merchants eligible for Agentic Storefronts. Google is rolling out its Universal Cart to unify agent-initiated purchases across its ecosystem. Tatcha, an early Shopify Agentic Storefronts adopter, reported 3x conversion rates and 11.4% of total store revenue from AI-assisted transactions.

The numbers behind the noise are real. AI-attributed orders on Shopify are up 11x since January 2025. Cyber Week 2025 generated $67 billion in AI-influenced sales. ChatGPT has 880 million monthly active users. The channel is material today. It is growing exponentially.

The Protocol Layer: How Agent-to-Merchant Transactions Actually Work

Agentic commerce required a new transaction architecture because agents do not use browsers. They call APIs. The Agentic Commerce Protocol (ACP) defines how four parties interact: buyers, agents, sellers, and payment providers.

The flow works like this. The agent interprets the buyer's intent and translates it into structured API calls. It creates a checkout session with the seller through REST endpoints. The seller calculates prices, taxes, shipping, and available inventory. The agent collects payment credentials through a Delegate Payment flow with the payment provider, which tokenizes them with spending limits. The seller processes the charge. The order is fulfilled.

Five REST endpoints handle the lifecycle: create a checkout session, retrieve state, update the session, complete with payment, cancel. Sellers remain the merchant of record. They own pricing, inventory, fulfillment, and payment processing. The protocol standardizes the conversation between agent and merchant so any agent can transact with any participating seller.

This matters because it means brands do not get to control the discovery experience. The agent decides which merchants to include in the checkout session. If the agent cannot find your product data, verify your entity, or match your inventory to the buyer's constraints, it calls a different seller's endpoint.

Why Traditional Analytics Cannot See Agentic Revenue

Here is the structural problem most brands have not caught yet. AI shopping agents bypass browser-based tracking entirely. No cookies. No JavaScript pixels. No click-path events. Server-side API calls generate zero client-side analytics. Your GA4 dashboard shows these orders as "direct" or attributes them to nothing at all.

This is not a tagging problem. It is an architectural gap. The Digital Applied attribution data shows AI chatbot referrals convert at 7% with a 38% average order value uplift, but most merchants cannot attribute those conversions because their measurement stack was built for a world where buyers use browsers.

The fix is server-side: API-level attribution, Shopify webhook integration, and checkout session correlation that ties agent-initiated transactions back to their source. Brands building this instrumentation now will understand the channel. Everyone else will keep making budget decisions on incomplete data while the channel grows behind their blind spot.

Agentic Commerce vs. Agentic Procurement

Agentic procurement is the B2B subset. An enterprise buyer tells an AI agent to find a Machine Relations agency with tier-1 media guarantees under $15K per month. The agent evaluates structured data, entity authority, cross-source corroboration, and citation frequency to build a vendor shortlist. Gartner projects 90% of B2B purchasing will be intermediated by AI agents by 2028.

Agentic commerce is the full paradigm. It covers consumer shopping (ChatGPT recommending and completing a headphone purchase), B2B vendor selection, multi-agent negotiation, and the protocol infrastructure connecting all of it. The AgenticPay benchmark models markets where buyer and seller agents negotiate through natural language across 110+ task configurations. The Agentic Commerce World (ACWorld) environment evaluates buyer and merchant agents across 200 tasks searching 785,022 transactable listings.

The academic and industry investment tells the story. This is not a feature inside existing commerce platforms. It is a new transaction layer being built from the protocol level up.

What This Means for Brand Visibility

Every Machine Relations principle compounds in agentic commerce. When the buyer is an algorithm:

Entity clarity determines whether the agent can identify your brand at all. Structured data, schema markup, and consistent entity signals across the web are table stakes for an agent conducting product discovery.

Citation architecture determines whether the agent trusts your brand enough to include it. Agents evaluate third-party coverage, publication authority, citation frequency, and recency. A brand with zero earned media presence in AI-retrievable sources does not make the shortlist.

AI extractability determines whether the agent can pull the data it needs to complete a transaction. Product specs, pricing, availability, and fulfillment options must be structured in formats an agent can parse and act on.

The brands that built AI visibility as a citation play are the ones positioned for agentic commerce. Citation was always the leading indicator of what mattered: machine-readable trust. Commerce is where that trust converts to revenue.

Frequently Asked Questions

What is agentic commerce?

Agentic commerce is the paradigm where AI agents autonomously research, compare, negotiate, and complete purchases on behalf of human buyers. The human sets the goal and authorizes the transaction. The agent handles everything in between: product discovery, price comparison, checkout, and payment through structured protocols like the Agentic Commerce Protocol.

How is agentic commerce different from AI-powered product recommendations?

Recommendations suggest. Agentic commerce acts. A recommendation engine shows you options you still have to evaluate, click through, and purchase yourself. An agentic commerce system takes your brief ("find noise-canceling headphones under $300, Sony or Bose, deliver by Friday"), executes the search, compares options, creates a checkout session with the merchant, and completes payment when you approve. The agent is the buyer. You are the decision authority.

Is agentic commerce happening now or is it theoretical?

Both. Forrester reports that most agentic experiences remain conversational in mid-2026, with true purchase autonomy still rare in consumer retail. But the infrastructure is live: Shopify Agentic Storefronts serve 5.6 million merchants, OpenAI takes a 4% cut on ChatGPT Shopping purchases, and Google is deploying Universal Cart. AI-attributed Shopify orders are up 11x since January 2025. The behavior is early. The revenue is real.

How do brands get included in agentic commerce transactions?

The same way they get cited by AI answer engines: through entity clarity, earned media authority, structured product data, and consistent cross-source corroboration. An AI shopping agent evaluates what it can verify. It does not respond to brand messaging or sales decks. It checks structured data, publication coverage, citation frequency, and entity signals. Machine Relations is the discipline that builds this presence.

Why can't my analytics see agentic commerce revenue?

AI shopping agents bypass browser-based tracking. They complete purchases through server-side API calls that generate zero cookies, zero pixel fires, and zero click-path events. Standard GA4 attributes this revenue to "direct" or misses it entirely. Server-side attribution, webhook integration, and checkout session correlation are required to see the channel. The attribution tracking fix covers the specific implementation.