What agentic commerce is

Traditional e-commerce assumes an active shopper: they open the site, type a query, browse the
catalogue, compare products and press buy. Agentic commerce changes that model: the person states a
task in natural language and an AI agent executes the whole chain.

Buy me the coffee beans I usually get, but with delivery by tomorrow — and the agent finds the
product, checks stock and a delivery slot, compares alternatives and places the order using the
shopper’s payment token.

The agentic commerce stack

Person
  ↓ a natural language request
AI agent (ChatGPT, Gemini and others)
  ↓ MCP / ACP / AP2
Store catalogue + order API
  ↓ shared payment token / network agent token
Order confirmation

Every layer needs standardisation: capability discovery protocols (MCP), ordering protocols (ACP),
payment mechanisms (tokens, mandates).

What changes for a retailer

Before Now
The storefront is the site or the app The storefront is the agent’s answer plus an API
SEO is the main discovery channel SEO plus catalogue agent-readiness
Conversion is a UX problem Conversion is a data and API quality problem
A funnel of four to seven steps A funnel of one or two (intent → purchase)

Important: in agentic commerce a store competes not for a position in search results but for
the agent choosing its catalogue — on data completeness, price, API availability and response
speed.

Where it stands in 2025–2026

The first working implementations are live: ChatGPT Instant Checkout (United States, Etsy and
Shopify merchants, from September 2025), with Google and other assistant vendors building comparable
transactional paths. Analysts project global agent purchase volume of 3 to 5 trillion dollars by
2030, conditional on broad adoption of the protocols and payment standards.

The practical priority for a retailer in the meantime is unglamorous: a structured catalogue, a
high-quality product feed, machine-readable product data, and an integration path to whichever
agent ecosystems matter in its markets.