How to Prepare for AI Agent Commerce

AI agent commerce is when autonomous agents complete purchases on a shopper's behalf, not just recommend products. The agent interprets intent, assembles a cart, and transacts through a payment protocol, which means your storefront needs a machine-readable and transactable data layer, not only a persuasive website.

This is the action end of the agent shift. Discovery gets you considered, but agent commerce requires that an agent can actually read your catalog, trust your prices, and pass a payment to your checkout without a human filling in a form.

The Agentic Commerce Protocol (ACP) is an open standard, maintained by OpenAI and Stripe and in beta as of 2026, that lets agents buy from you while you keep control. A transaction runs in four steps that preserve your existing payment relationship.

First, the agent assembles a cart from your product feed. Second, the buyer selects a payment method inside the agent surface. Third, the payment provider issues a scoped token. Fourth, the agent posts that token to your checkout endpoint, and you charge through your own payment service provider. Settlement and disputes stay on your side, and one integration can open selling across multiple agents.

Google's Agent Payments Protocol (AP2) is a complementary standard focused on authorizing agent payments with a verifiable audit trail. Announced September 16, 2025 with more than 60 launch partners including Mastercard, PayPal, Coinbase, and American Express, it reached v0.2.0 in April 2026 and was donated to the FIDO Alliance.

AP2 introduces three signed mandates carried as W3C Verifiable Credentials. The Intent Mandate captures the shopper's instruction, such as buy running shoes under 120 dollars. The Cart Mandate records exactly what is bought at what price. The Payment Mandate authorizes the charge, and because AP2 is payment-method agnostic it spans cards, bank transfers, real-time payments, and stablecoins.

Every protocol depends on the same foundation: a complete, accurate, machine-readable product feed. If an agent cannot parse your catalog or trust your prices, no checkout standard will save the sale.

Prioritize the data layer before the plumbing. Aim for real-time price and inventory accuracy, complete attributes, and consistent taxonomies, then add Product and Offer schema in JSON-LD so your on-page data corroborates the feed an agent reads.

A key reason to adopt these standards is that they are designed to keep merchants in control. Under ACP, a single integration lets you sell through agents while retaining control over what is sold, how your brand appears, and how orders are fulfilled.

Crucially, you charge through your own payment provider, and settlement and dispute handling stay merchant-side. ACP is also designed to work even if you do not process with Stripe, so you can adopt it with your existing payment stack rather than re-platforming.

The near-term reality is discover in AI, buy on your site, even as full agent checkout matures. In-chat checkout stalled in early 2026 after merchants saw it convert worse than click-throughs, so the safest bet is to be excellent at discovery now while getting protocol-ready for transactable commerce.

Because these standards are young and evolving, hedge your roadmap: build the machine-readable data foundation that pays off regardless of which protocol wins, and pilot ACP or AP2 support once your feed, pricing accuracy, and checkout endpoint are solid.