Getting Started with Agentic Commerce Protocol: A Complete Guide for Mid-Market Merchants

Last week, a customer asked ChatGPT "What's the best turbocharger for my 2015 WRX?" Within seconds, they were browsing options, comparing specs, and completing a purchase—all without leaving the chat. This is agentic commerce in action, and it represents a fundamental shift in how 700 million weekly ChatGPT users discover and buy products.

For mid-market merchants, this isn't about adopting another sales channel. It's about being discoverable in a new way customers are already shopping.

ACP in 60 Seconds

The Agentic Commerce Protocol (ACP) is an open standard developed by OpenAI and Stripe that enables AI agents to help users discover and purchase products directly within conversations. It powers "Instant Checkout" in ChatGPT, where customers can complete purchases without switching to a browser.

Think of ACP as the technical bridge between AI agents and your e-commerce system. It defines how products are discovered, how shopping carts work, how payments process, and how orders flow into your fulfillment systems—all triggered by natural language conversations.

What makes ACP unique:

You remain the merchant of record. You own the customer relationship, control pricing, and manage the post-purchase experience. Payment flows directly to your bank account through your existing processor. The AI agent is simply a discovery channel—like Google Shopping or Facebook Marketplace, but conversational.

Why This Matters for Your Business

The discovery opportunity: When someone asks an AI agent for product recommendations, they're showing high purchase intent. They're not browsing—they're ready to buy and need help making the right choice. Being part of that conversation means capturing customers at a critical decision moment.

The economic shift: AI-driven commerce introduces performance-based economics. Instead of paying for impressions or clicks, you pay a percentage when transactions complete. The costs align directly with revenue, making ROI measurement straightforward.

The platform advantage: If you're on Shopify, BigCommerce, or Salesforce, ACP support is built-in. But if you run Adobe Commerce (Magento), Miva, NetSuite SuiteCommerce, or custom platforms, you'll need additional infrastructure. Understanding this gap now gives you time to plan implementation thoughtfully.

Early adopters gain valuable insights into how customers interact with AI agents, positioning themselves with optimization advantages as the channel matures.

The Three Components of ACP

ACP has three core components:

1. Product Feeds

This is how you communicate product information to AI agents. Unlike traditional feeds optimized for visual search, ACP requires structured, machine-readable data. AI agents evaluate products based on specifications, certifications, compatibility data, and quality indicators.

The real work is restructuring your catalog. Most e-commerce systems optimize for human browsing. Descriptions like "Premium quality blue shirt" work when customers see photos. For AI agents making comparisons, you need "100% Giza 87 Egyptian cotton, single-needle stitching, Fair Trade certified, lifetime warranty."

2. Checkout APIs

These endpoints enable AI agents to create shopping carts, calculate shipping and taxes, and complete purchases. When a user says "Add that to my cart," the agent calls your API to create a session, calculate pricing, and present fulfillment options.

Performance requirement: sub-300ms response times. Many legacy platforms need optimization through edge caching or microservices architecture.

3. Delegated Payments

This security mechanism allows payment credentials to flow safely. Users authorize OpenAI to spend up to a certain amount. OpenAI works with Stripe to generate single-use tokens. Your system charges these tokens using your existing payment setup. Money flows directly to your account—OpenAI never becomes the merchant of record.

The Real Challenge: Your Data Isn't AI-Ready

The hardest part of ACP implementation isn't building APIs—it's restructuring product data so AI agents can understand it.

Automotive aftermarket example:
Instead of: "Premium electrical wire"
AI needs: "12 AWG THHN/THWN-2 copper wire, 600V rated, UL listed, suitable for conduit installations in commercial buildings per NEC Article 310"

This semantic translation from marketing copy to machine-readable specifications takes effort, but it's also your competitive advantage. Products with better data structure get recommended more often.

Three Paths to ACP Compliance

You have three paths to ACP compliance:

Build in-house: 500-800 developer hours, $75K-$200K, 6-12 months timeline. You own the code but also maintain it as ACP evolves. Best for large enterprises with dedicated platform teams.

Hire a systems integrator: $100K-$250K, 6-12 months, custom build with code ownership transferring to you. Good if you have internal development capacity but lack ACP expertise.

Use specialized middleware: Managed services handle ACP compliance as infrastructure, similar to using Stripe for payments. Faster implementation (weeks vs months), vendors manage protocol updates. Ideal for merchants who want to focus on their core business.

The strategic question: Is maintaining AI commerce infrastructure your highest-value use of engineering resources?

ACP Isn't the Only Standard (Plan Accordingly)

ACP isn't the only standard. Google, Shopify, and Walmart support the Universal Commerce Protocol (UCP). Anthropic uses the Model Context Protocol (MCP) for Claude integration. More standards will emerge as platforms compete.

Building separate integrations for each protocol can be resource-intensive. Consider whether you need abstraction layers that handle multiple standards, giving you flexibility as the ecosystem evolves.

Four Signs You're ACP-Ready

From a business perspective, successful ACP implementation means:

Visibility in AI recommendations: When customers ask AI agents for products in your category, your products appear in their consideration set.

Seamless order flow: AI agent orders integrate smoothly into your existing order management and fulfillment workflows. Your team processes them like orders from any other channel.

Performance reliability: AI agents complete purchases without latency issues that would cause them to recommend competitors instead.

Data sovereignty: You maintain complete control over customer information, transaction data, and post-purchase relationships.

The Window Is Open. Early Adopters Win.

The opportunity with agentic commerce is accessibility. Early implementation gives you time to optimize product data, understand customer behavior patterns in AI conversations, and refine your approach before the channel becomes highly competitive.

Adobe Analytics reports strong year-over-year growth in AI-referred traffic. Salesforce data shows increasing influence of AI agents on purchasing decisions. The question isn't whether agentic commerce will matter—it's how you want to participate.