2026代理式电商实战手册：驾驭产品发现标准

Navigating the Standards of EXECUTIVESUMMARY Commerce hasalready changed. AI already controls product discovery.Millions of people no longer browse;37% ofconsumersnow start their searches on The only friction left is the checkoutprocess, and the hold-up there is in Today,58% of consumersuse AI forresearch, yet only 17% trust it to purchase.That gap will close, and when it does,brands will face a binary choice: Be agent-readable, or become invisible. Protocols like MCP, A2A, UCP, and ACP aredefining how this new ecosystem operates.The winning standard does not matter as One principle is absolute: AIdoes not interpret your brand, If your product information is fragmented,inconsistent, or incomplete, agents will bypass you. If your data is structured, enriched, andinteroperable, agents will prioritize you. Agentic commerce will not replace eCommerce, itwill dominate as an entirely new channel. The question is simple: WhenAI becomes the buyer, will it FOREWORD BENOIT JACQUEMONT,CTO & CO-FOUNDER Since starting my eCommerce career in 2005, I’veseenseveral major shifts reshape the digitalcommerce industry: the rise of eCommerce, themove to mobile, the dominance of marketplaces,the emergence of composable architectures, and Thenext step is inevitable:AI will not onlyrecommend products, but will also autonomously This evolution raises important questions abouttrust, infrastructure, standards, and control. It alsoreinforces something I have believed throughoutmy career: commerce runs on data. No matter theinterface, be it website, mobile app, marketplace, Agenticcommerce is another one of those What makes this shift different is not just thetechnology, but the rapid pace of growth and theassociatedbehavioral changes.For years,weoptimizedforkeywords,navigation,andconversion funnels. We are now moving to intent- Agenticcommerce will not replace everythingthat came before it. But it will add a powerful newlayerto how products are discovered andpurchased. The businesses that prepare now by KEY TERMS Agentic Commerce 1 A form of digital commerce where autonomous AI agents independently discover products,make purchasing decisions, and complete orders on behalf of users, minimizing direct human Agentic Commerce Protocol (ACP) An OpenAI- and Stripe-led commerce standard that enables product discovery and checkoutinside controlled agent platforms such as ChatGPT and Microsoft Copilot. Merchants provideproduct data through a feed-based model rather than being openly discoverable across the 2 Agent-to-Agent (A2A) 3 An open protocol that enables secure communication, task delegation, and collaborationbetween multiple AI agents so they can coordinate and delegate activities during complex Agent Payments Protocol (AP2) 4 A protocol that standardizes secure payment tokenization, allowing AI agents to execute paymentson behalf of users without exposing sensitive financial data to the agents and merchants. AI-Augmented Commerce The current phase of eCommerce where AI tools help consumers discover products, refinesearch intents, and make decisions more efficiently, but still require users to complete checkout 5 “Buy For Me” Experience 6 A seamless shopping experience where an AI agent autonomously executes a purchase on theuser’s behalf, from product discovery through checkout, without the user performing each step. Manifest File A machine-readable file associated with UCP that is intended to describe a merchant’scommerce capabilities and how AI agents can interact with them. However, product discovery inUCP today is still primarily feed-based via Google Merchant Center; manifest-driven discovery 7 Model Context Protocol (MCP) 8 An open-source standard introduced by Anthropic that gives AI agents a common way toaccess and interact with real-time data and tools across systems, though it does not itself Universal Commerce Protocol (UCP) A Google- and Shopify-backed open standard designed for agentic commerce across the openweb. Its long-term vision is manifest-driven, real-time agent discovery and interaction, but today 9 THE PATH TOAGENTICCOMMERCE The Past: Traditional eCommerce Traditional eCommerce follows a structured, step-by-step journey where the consumer is responsiblefor every action.While familiar, it requires time, comparison, and constant navigation between When a consumer has an intent, for example, renovating their coffee table, they must translate thatintent into a series of actions: The Present: “AI-Augmented” eCommerce By the end of the conversation, the user has acurated list of products.However, the flow breaks AI assists, humans still do the work. Instead of conducting research across multiplesites, the consumer simply enters their intent:"I Once the list is generated, the consumer must stillleave the chatbot and visit one or multiple websites The AI chatbot responds by: Clarifying the type of woodAsking about the desired finishSuggesting the correct grit prog