AI shopping agents, once a combination of novelty and hype, are becoming a viable ecommerce sales channel.
Last month, Google and Stripe announced new shopping and payment processing tools for AI-related purchases.
Standards
Google’s Agent Payments Protocol (AP2) and Stripe’s Agentic Commerce Protocol (ACP) offer a degree of transaction standards for agent-based commerce.
That standard addresses concerns about turning AI-driven discovery into sales while keeping payments safe and authorized.
For example, how does a merchant know if a human authorized an AI agent to make a purchase? Who is responsible if the buyer contests it?
Once a novelty, AI shopping agents are becoming a reality for consumers.
Google’s AP2
AP2 seeks to answer those questions. The protocol provides a framework for authenticating and validating AI-led transactions, enabling banks, merchants, and consumers to trust the outcome.
The solution uses what Google called “tamper-proof, cryptographically-signed digital” mandates with verifiable credentials.
A mandate is a digital contract between the shopper and her AI agent. It proves what she asked the agent to do and under what conditions. That mandate then becomes the evidence that merchants and payment providers can rely on if she disputes the transaction.
AP2 documents the mandate when, for instance, a shopper tells an AI interface to buy a pair of blue, size 10 men’s running shoes. Thus AP2 could be a foundation for agentic commerce that payment networks, regulators, and retailers can accept as legitimate.
Stripe’s ACP
While AP2 focuses on trust and authorization, Stripe’s Agentic Commerce Protocol focuses on turning conversations into transactions. ACP provides merchants with a way to present products, pricing, and checkout information in a format usable by AI agents.
Behind the scenes, Stripe issues a Shared Payment Token that passes payment details from the AI agent to the merchant without exposing the shopper’s credentials. The order itself flows into the merchant’s backend via ACP to be accepted and fulfilled, like any other ecommerce order, which the buyer can return later.
Stripe launched ACP alongside its Instant Checkout in ChatGPT. U.S.-based Etsy sellers are the first participants, with Shopify merchants expected to follow. A shopper asks ChatGPT for recommendations, chooses a product, and completes payment through a Stripe-powered checkout without leaving the chat.
Implications
New technologies often create both opportunities and challenges. For merchants, agentic commerce is no exception.
A promising opportunity may lie in latent demand. Shoppers sometimes want to buy a product, but the item is out of stock, too expensive, or takes too long to ship. A simple search doesn’t result in a sale.
An AI shopping agent could scour dozens or even hundreds of merchant sites and feeds, identifying the proper product at the right price, which the consumer would have otherwise missed. This solves the latent demand.
Yet instantly comparing prices, shipping rates, and product availability across many merchants creates intense competition. That’s the challenge.
It could even lead to platform pressure, where merchants are present on multiple channels — Etsy, Shopify, Amazon, eBay, and others — to access the dozens of AI shopping portals.
Permission Marketing
Imagine a shopper prompting an AI agent to “look for noise-canceling headphones less than $200, and buy them if they go on sale during Black Friday.”
The agent monitors products, validates prices, and executes purchases automatically. The scenario — an AI agent shopping for a human — is futuristic, but is also familiar to those old enough.
In his 1999 book, “Permission Marketing: Turning Strangers into Friends, and Friends into Customers,” Seth Godin argued that interruption marketing — television commercials, display ads, email promos — was losing its effectiveness. Businesses needed consumer consent to build ongoing relationships. Permission, not promotion, was the key.
Shoppers who allowed merchants to execute purchases on their behalf were Godin’s ultimate example of trust.
Agentic Permission
In all likelihood, agentic commerce will depend on the same foundation Godin described a generation ago: trust and consent.
The key will be how ecommerce businesses react when algorithms programmatically select and buy products.
Permission marketing, such as opt-in email, forced merchants to rethink customer relationships. Similarly, agentic commerce could require merchants to prove reliability, value, and transparency.
Success may hinge less on who has the flashiest storefront and more on which merchants provide the cleanest data, fairest pricing, and the most dependable fulfillment for AI agents to recognize.