The Rise of Agentic Commerce and the Fundamental Transformation of the Digital Checkout Experience
The digital storefront, a staple of global commerce for over three decades, is undergoing its most significant architectural shift since the invention of the Secure Sockets Layer (SSL) protocol. For thirty years, the "checkout" has been defined as a destination—a specific webpage where consumers manually input names, addresses, and credit card numbers. While innovations like Amazon’s one-click patent and Apple Pay’s biometric authentication reduced the time spent on these forms, the form itself remained the central bottleneck of online trade. That paradigm is now collapsing as the industry moves toward "agentic commerce," a model where checkout is no longer a page to be visited but a protocol to be executed by artificial intelligence.
This transition reached a critical inflection point between late 2025 and early 2026 with the launch of two competing open standards: the Agentic Commerce Protocol (ACP), spearheaded by Stripe and OpenAI, and the Universal Commerce Protocol (UCP), unveiled by Shopify and Google. These frameworks represent a future where AI agents—autonomous software capable of reasoning and taking action—handle the entire purchasing journey, from product discovery to final payment, without a human ever loading a merchant’s website.
A Chronology of Friction: The Road to Autonomy
To understand the magnitude of agentic commerce, one must view it as the logical conclusion of a thirty-year effort to remove friction from the "intent-to-ownership" pipeline. Each generation of commerce technology has solved a specific logistical or cognitive hurdle.
The era of online commerce began on August 11, 1994, when Phil Brandenberger purchased a Sting CD from the website NetMarket for $12.48. This transaction proved that encryption could protect financial data, removing the friction of physical distance. By the late 1990s, the rise of comparison-shopping engines like BizRate and PriceGrabber addressed the friction of information asymmetry, allowing users to compare prices across multiple vendors without visiting each store.
In 1998, Amazon revolutionized the discovery phase by deploying item-to-item collaborative filtering. This algorithm, which powered the "customers who bought this also bought" feature, removed the friction of choice by predicting consumer needs before they were explicitly stated. The 2010s saw the rise of conversational and social commerce, led by platforms like WeChat in China and later TikTok Shop in the West. These platforms moved the storefront into the social feed, removing the need to navigate away from entertainment to make a purchase.
By 2024, AI shopping assistants like Amazon’s Rufus and Google’s AI-enhanced Shopping Graph began doing the research, comparison, and recommendation work for the user. The final leap occurred in 2025 with the introduction of OpenAI’s "Operator" and Google’s "Buy for Me" features. These tools removed the final remaining friction: the human presence. In this new era, the transaction happens via API calls between an AI agent and a merchant server, rendering the traditional user interface obsolete.
The Architecture of Invisible Checkout: ACP vs. UCP
The shift from a "page-based" checkout to a "protocol-based" checkout is currently being defined by two primary frameworks that, while different in scope, share a common goal of making commerce machine-readable.
The Agentic Commerce Protocol (ACP)
Announced in September 2025, ACP is a collaboration between OpenAI and Stripe. It focuses heavily on the transactional layer—the moment of purchase. ACP utilizes a four-party model involving the buyer, the AI agent, the merchant, and the payment service provider. Under this protocol, the merchant remains the "merchant of record," but the AI agent handles the user experience.
The protocol defines four essential API endpoints:
- Create Checkout: The agent sends a product SKU, and the merchant generates a cart with real-time pricing and shipping options.
- Update Checkout: Allows for mid-flow modifications, such as changing shipping methods.
- Complete Checkout: The agent sends a secure payment token, and the merchant processes the payment.
- Cancel Checkout: Signals the release of reserved inventory.
Stripe’s Agentic Commerce Suite, launched in late 2025, provides a low-code gateway for businesses to adopt ACP, allowing them to sell across multiple AI platforms—including Microsoft Copilot and Anthropic’s Claude—through a single integration.
The Universal Commerce Protocol (UCP)
Launched in January 2026 by Shopify and Google, UCP takes a broader approach. While ACP focuses on the transaction, UCP is designed as a full-stack commerce standard modeled after the TCP/IP architecture of the internet. It includes layers for discovery, capabilities (such as catalog management), and domain-specific extensions.
UCP is protocol-agnostic, supporting REST, MCP (Model Context Protocol), and Google’s own AP2 (Agent Payments Protocol). It relies on a "well-known" endpoint (/.well-known/ucp) where merchants publish their capabilities. This allows AI agents to "handshake" with a merchant’s server to determine what actions they can perform together, such as checking local inventory or processing a complex return.
The "Person-Not-Present" Challenge and Payment Security
The most significant hurdle for agentic commerce is trust. Traditional fraud detection relies on human behavioral signals—how a user moves their mouse, their typing speed, or their browsing history. AI agents exhibit none of these behaviors. This has given rise to a new category of security concern: "person-not-present" transactions.
To address this, the financial industry has introduced several new primitives:
- Shared Payment Tokens (SPTs): Developed by Stripe, these are programmable tokens scoped by merchant, time, and amount. The buyer’s actual credit card details are never exposed to the agent or the merchant.
- Trusted Agent Protocol: Introduced by Visa in October 2025, this framework uses HTTP Message Signatures to help merchants distinguish legitimate AI agents from malicious bots.
- Agent Pay: Mastercard’s solution involves "Agentic Tokens" that build on existing tokenization infrastructure, allowing consumers to set strict spending limits and permissions for their AI assistants.
- AP2 (Agent Payments Protocol): Google’s standard uses cryptographic mandates to provide tamper-evident proof of user consent, ensuring that an agent cannot exceed its authorized purchasing power.
Despite these safeguards, the attack surface for agentic commerce is novel. Research from 2025 and 2026 has highlighted the risk of "visual prompt injection," where malicious code embedded in product images or reviews can hijack an AI agent’s logic, potentially causing it to redirect funds or purchase incorrect items.
Market Projections and Early Adoption
The economic implications of this shift are staggering. McKinsey & Company projects that by 2030, agents will orchestrate approximately $1 trillion in U.S. retail revenue, with global figures reaching up to $5 trillion. Gartner offers an even more aggressive forecast for the B2B sector, predicting that 90% of B2B purchases will be handled by AI agents by 2028, representing $15 trillion in spending.
Consumer adoption is following a generational divide. While a Contentsquare survey found that only 30% of the general U.S. population is currently willing to let an AI complete a purchase, that number is significantly higher among Gen Z and early adopters. Data from Adobe Analytics shows that AI-driven traffic to retail sites grew by 4,700% year-over-year by mid-2025. Shopify has reported that orders attributed to AI-driven searches have increased 11-fold since the start of 2025.
Early adopters of these protocols include a diverse array of global brands. Retail giants like Walmart and Instacart are already live with ChatGPT’s Instant Checkout. Fashion and lifestyle brands such as Glossier, SKIMS, and Urban Outfitters have onboarded to Stripe’s agentic suite, while Gymshark and Everlane are utilizing Google’s AI Mode via UCP.
Implications for Businesses: The Shift to Machine-Readability
For merchants, the rise of agentic commerce necessitates a fundamental shift in digital strategy. In the traditional era, businesses optimized for "Conversion Rate Optimization" (CRO), focusing on button colors, layout, and human psychology. In the agentic era, the focus shifts to "Agentic AI Optimization" (AAIO).
To remain competitive, businesses must ensure their product data is impeccably structured. AI agents do not "look" at websites; they parse them. This means that high-quality, high-resolution images with accurate alt-text, precise SKU data, and detailed attribute mapping (weight, dimensions, materials) are no longer optional.
Furthermore, the adoption of Schema.org markup is becoming a critical trust signal. Experts suggest that consistent, accurate structured data builds "machine comfort bias," where AI systems prioritize merchants that provide reliable, easy-to-parse information over those with opaque or poorly formatted data.
A New Frontier in Global Trade
The transition to agentic commerce represents the final stage of the "invisible" transaction. As checkout evolves from a physical page into a background protocol, the relationship between brands and consumers will be mediated by intelligent intermediaries.
Kevin Miller, Stripe’s Head of Payments, noted that the company spent fifteen years optimizing commerce for humans, and it is now embarking on the same journey for agents. For businesses, the message is clear: the agents are already shopping. Whether they can find and interact with a specific store depends entirely on that store’s willingness to speak the language of the machines. The era of the "checkout page" is ending, replaced by a seamless, automated flow that promises to redefine the global economy.
