What happened to agentic commerce, by date?
A year ago, many expected shoppers to stop visiting websites and marketplaces. The timeline shows a slower path.
- September 2025: OpenAI and Stripe launch Instant Checkout, which lets users buy inside ChatGPT.
- December 2025: Visa reports hundreds of agent transactions in pilots.
- January 2026: Google presents the Universal Commerce Protocol with Etsy, Shopify, Targettarget variable. A target variable is the value a machine learning model learns to predict: whether a customer will leave, how much product will sell, how much a flat costs., Wayfair, and Walmart. Alibaba builds Alipay into the Qwen app.
- February 2026: Alipay’s AI payment handles 120 million transactions in one week. ChatGPT has 900 million weekly users.
- March 2026: OpenAI closes Instant Checkout and moves to store apps inside the chat. Stripe admits that agentic commerce was hyped too early.
Compare the scale: hundreds of pilot transactions at Visa against 900 million weekly ChatGPT users. China is the exception, because Alipay’s AI payment works inside the Qwen app.
Why did shoppers not hand purchases to agents?
Shoppers want advice from AI but prefer to pay themselves. Forrester estimates that few buyers use agent purchases, and that the share barely changes, although interest grows slowly.
People pay on a familiar platform because of its returns and buyer protection. The Chinese example works because payment, chat, and store sit in one app with a shared habit. Elsewhere, a buyer has no reason to change route when the seller answers fast and clearly where the buyer already is.
What brings sales to a marketplace seller today?
A seller earns money today from quick answers, not from an agent that buys for the customer. An agent that replies to buyers within a minute in the marketplace chat, in messengers, and on the website wins the order before a competitor answers the next morning.
A seller-side AI assistantAI assistant. An AI assistant is a helper built on a neural network that, at a person’s request, answers questions, prepares texts and does simple tasks in work programs. can do five jobs:
- Answer buyers about stock, delivery, and returns.
- Reply to reviews and questions on product pages.
- Pass disputed cases to a person at once, so no deadline is missed.
- Log every contact in the CRM with dates.
- Work only through the official tools of each platform.
Our pilot takes 14 days on one platform, with the metric written into the contract: response time and the share of questions closed without a person. The price and the term are fixed in the contract.
What does this mean in Russia?
Russian marketplaces and messengers go the same way: AI assistants advise, but keep purchase and payment to themselves. For a seller on Russian platforms, the money in the coming years comes not from an “agent that buys for the customer” but from an agent that answers a buyer within a minute in the platform chat, on Avito, in Telegram and on the site, while a competitor answers in the morning.
From 1 October 2026 the platform economy law (289-FZ) also makes correspondence in the seller’s account legally significant, and the platform must consider a seller’s complaint within 15 days. Speed and completeness of answers have become both sales and protection.
How can you prepare for agentic commerce now?
Prepare three things. They cost little and help in any market scenario.
- Complete product cards: composition, sizes, terms, and delivery times.
- One list of answers to buyer questions. Today your manager reads it. Tomorrow your agent and the buyer’s agent will read it.
- One person responsible for marketplace messages.
A knowledge baseAI knowledge base. An AI knowledge base is a set of selected and prepared company documents that an AI assistant searches for answers: rules, instructions, a product catalog and answers to common questions. is a set of selected company documents that an AI assistant searches for answers. A seller with a ready knowledge base will earn first when agent purchases arrive, because buyer agents need the same accurate product data.