Skip to content

Agentic commerce

| Article: 5 min read | AI in e-commerce

Agentic commerce is shopping in which an AI agent acts on a person's behalf: it searches for and compares products, and within the permission it has been given it can also place the order and pay. For a store, this means part of the "visitors" will be programs rather than people, reading your catalog, prices and terms.

In plain words

It used to be that a person searched Google, opened sites, compared and paid. Now they can ask an AI assistant: "find running shoes under 3,000 UAH with Nova Poshta delivery." The assistant browses stores, compares and returns a shortlist. In a more advanced version it also completes the purchase with the person's approval. For what an agent is, see AI agents.

What has already launched

Status as of November 2026, from public sources. Everything below mostly applies to the US for now.

  • Agentic Commerce Protocol (ACP) from OpenAI. The OpenAI developer page describes three parts: product feed delivery (so ChatGPT sees current price and availability), a checkout specification, and a "delegated payment" specification for payment providers (OpenAI Developers).
  • Instant Checkout in ChatGPT was shut down. According to The Keyword on March 25, 2026, OpenAI stopped purchases without leaving the chat: after searching, the user now goes to checkout on the retailer's site. The reasons given in the same piece: only about 12 of the millions of Shopify merchants launched the feature, shoppers preferred sites where their accounts and cards were already saved, and syncing data across millions of catalogs proved hard. Product search and comparison tools remain, and the retailers Target, Sephora, Nordstrom, Lowe's, Best Buy, The Home Depot and Wayfair work through ACP (The Keyword, 2026). In other words, the first large-scale experiment with buying inside a chatbot did not work out.
  • Universal Commerce Protocol (UCP) from Google. An open standard announced on January 11, 2026, built together with Shopify, Etsy, Wayfair, Target and Walmart; more than 20 other companies, including Visa, Mastercard, Stripe and Adyen, support it. A store describes its capabilities in a JSON manifest at `/.well-known/ucp`, and the agent creates and updates checkout sessions. According to Google, the seller remains the "merchant of record," the party that sells and is responsible for the order (Google Developers Blog, 2026).
  • Agent Payments Protocol (AP2) from Google, announced on September 16, 2025. It works through "mandates": signed digital documents that record exactly what the person has authorized (budget limits, a specific cart, a payment method) (AP2).
  • Agentic purchase in Google Search. Since November 13, 2025, in the US, Google Search and AI Mode let you ask Google to buy a product on a merchant's site through Google Pay. Google says it will "always ask for permission first," and the user confirms the purchase and delivery. Among the first merchants are Wayfair, Chewy, Quince and some Shopify stores (TechCrunch, 2025).

There is no information about a launch in Ukraine or Europe.

What the data on shoppers says

  • According to Adobe, in the 2025 holiday season traffic to US retail sites from generative AI services grew 769% in November and 673% in December year over year; these visitors converted 31% better than others (Digital Commerce 360, 2026). But Adobe gives neither the starting volume nor AI's share of total traffic, so the percentages do not show real scale. Practical Ecommerce concludes that AI did not define the season, though it is worth tracking (Practical Ecommerce, 2026).
  • Accenture survey (25,590 consumers in 16 countries, January 2026): 32% would let an agent choose a purchase within set limits if a person confirms the payment; fully autonomous purchases are acceptable to only 9%. These are survey answers, not actual behavior (Accenture, 2026).
  • According to Gartner (2026), customers were three times more likely to use third-party AI services (ChatGPT, Gemini, Copilot) than company chatbots (Gartner via DestinationCRM, 2026). This is about service requests, but it shows that people are already talking to a store through an intermediary.

What this means for a store

What follows are conclusions from the facts above, not a forecast.

  1. Product data becomes the storefront. An agent reads the feed, the structured data and the product page. Complete specifications, current price and availability, and clear delivery and return terms help both people and agents. See product feed, PIM, product page.
  2. Your site remains the place of payment. After Instant Checkout was shut down, even in the US the purchase mostly happens on the seller's site. A good checkout does not lose its value.
  3. Track the new traffic source. In analytics, look at visits from AI services separately from Google and social networks.
  4. Ukrainian realities. Cash on delivery, Nova Poshta and payments in hryvnia need a separate solution in any protocol. There is no fresh data on how this works with agents.

Common mistakes

  • Waiting for "in-chat purchases" as a mass channel: the first big launch was wound down.
  • Taking growth percentages for volume: 700% of a small number is still a small number.
  • Plugging into protocols before the catalog is in order.
  • Not thinking through who is responsible for an agent's mistake (wrong size, address, an extra order).

FAQ

What is agentic commerce in plain words? It is when an AI assistant chooses, and sometimes completes, a purchase on a person's behalf, within the permission the person has given.

Can AI agents already buy products on their own? In the US, yes, in limited scenarios: for example, an agentic purchase in Google Search with user confirmation. There is no mass autonomous shopping, and chat purchases in ChatGPT were shut down in March 2026.

How is ACP different from UCP? ACP is OpenAI's protocol, UCP is Google's open standard with partners. Both describe how an agent sees the catalog and goes through checkout. They are parallel standards, and there is currently no single one that everyone supports.

Does a Ukrainian store need to connect to these protocols? There is no public data on availability in Ukraine. What is already useful in practice: a complete feed, accurate specifications, clear delivery and return terms.

Can you trust an agent with payment? Accenture's data shows caution: only 9% of respondents allowed full autonomy.

Sources

Want to check this on your store?

AI automation

An AI assistant checks the store every day, suggests fixes and makes them after your «Yes».

About the service →

Share

Related

  • Technical debt

    | Article: 6 min read

    Technical debt is the future cost of fixing and reworking that arises when software was built fast and without room for change. Like financial debt, it lets you move faster today, but with interest: every next change costs more and takes longer.

  • Vibe coding

    | Article: 4 min read

    Vibe coding is building programs where a person describes the task in ordinary language, AI writes the code, and the person accepts the result almost without checking it and goes by whether it "looks like it works", adding new wishes in the chat. For quickly…

  • How an online store team uses AI every day

    | Article: 2 min read

    AI in a store team's daily work is a set of small, repeating tasks in which a language model prepares a draft and a person checks it and decides. The value comes not from separate "smart" features but from simple habits assigned to specific people.

  • AI automation for an online store

    | Article: 5 min read

    AI automation for an online store means handing part of the repetitive work to a language model (ChatGPT, Claude, Gemini and the like): draft descriptions and translations, a first reply to a customer, search through a knowledge base, spreadsheet analysis. It…