What Is Agentic Commerce?

How Agentic Commerce Works

In traditional e-commerce, a person does everything: they search for a product, read reviews, compare prices across multiple tabs, add an item to a basket, enter their payment details, and confirm the order. Agentic commerce shifts part or all of that process to an AI agent.

A simple example: you tell an AI assistant to find running shoes under £80 with good arch support and next-day delivery. The agent searches multiple retailers, compares specifications, checks stock availability, and presents the best options. In more advanced implementations, the agent completes the purchase without you clicking anything at all.

The key distinction is autonomy. A chatbot that recommends a product but waits for you to press "buy" is conversational commerce. An agent that completes the entire purchase within agreed parameters is agentic commerce.

Why It Matters Now

Several things have converged in 2025 and early 2026 to move agentic commerce from concept to reality.

First, the AI models have become capable enough to handle multi-step tasks reliably. Modern agents can reason through a shopping decision, call external APIs, compare structured product data, and execute a checkout flow.

Second, the infrastructure now exists. In September 2025, OpenAI and Stripe launched the Agentic Commerce Protocol, an open standard that lets AI agents complete purchases inside ChatGPT. In January 2026, Google released the Universal Commerce Protocol at the National Retail Federation conference, co-developed with Shopify, Etsy, Wayfair, Target, and Walmart, with endorsement from over 20 global partners including Visa, Mastercard, and American Express.

Third, consumer behaviour is already shifting. An IBM Institute for Business Value study found that 45% of consumers already use AI for part of the buying journey. Half of all consumers now use AI when searching the internet, according to McKinsey.

Levels of Automation

Not all agentic commerce looks the same. McKinsey describes an "automation curve" with distinct levels.

At the lowest level, automation handles simple recurring purchases like subscription refills. This is what services like Amazon Subscribe and Save already do.

At the next level, agents assist with one-off purchases. You describe what you want, the agent finds options and presents them, but you still confirm the final purchase.

At higher levels, agents operate against standing goals rather than single transactions. For example, "keep household essentials under £250 per month" or "maintain my airline loyalty status at the lowest cost". The agent continuously monitors, anticipates needs, and acts accordingly.

Higher automation is not always better. The goal is not maximum autonomy but optimal delegation. People will happily automate the reorder of washing powder. They are far less likely to delegate the purchase of a holiday or a car.

What This Means for Businesses

For businesses, agentic commerce changes the fundamental question of who the customer is. Your next customer might not be a person browsing a website. It might be an AI agent evaluating your product data, checking your stock levels, and completing a purchase on someone's behalf.

This has implications for product data quality, API accessibility, and, critically, for identity and compliance. When an agent makes a purchase, the business still carries legal liability. Someone needs to prove that the person behind the agent authorised the transaction, that spending limits were respected, and that regulatory requirements were met. This is where compliance infrastructure becomes essential.

Frequently Asked Questions

What does agentic commerce mean?

Agentic commerce means AI agents acting autonomously to research, compare, and purchase products or services on behalf of a human or business. The word "agentic" comes from "agency", meaning the capacity to act independently within defined parameters.

How is agentic commerce different from conversational commerce?

In conversational commerce, you chat with an AI that suggests products, but you still make the final purchasing decision. In agentic commerce, the AI agent can execute the purchase autonomously within the limits you have set. The simplest test: if the human clicks the buy button, it is conversational commerce. If the AI does, it is agentic commerce.

Is agentic commerce already live?

Yes. As of early 2026, ChatGPT users in the US can purchase from Etsy sellers directly within the chat, with over a million Shopify merchants coming soon. Google has announced native checkout in AI Mode in Search and the Gemini app through the Universal Commerce Protocol. Mastercard has launched Agent Pay technology for AI agent transactions.

What are the risks of agentic commerce?

The primary risks are around identity, liability, and fraud. When software initiates a purchase rather than a person, businesses need to prove who authorised the transaction, what limits applied, and whether regulatory requirements were met. Existing legal frameworks were designed for human decision-makers, creating gaps around consent, contract formation, and dispute resolution that the industry is actively working to close.