Agentic commerce for retail: being buyable inside the assistant

Retail is the category where conversational discovery bites first. Shoppers ask an assistant for a winter jacket under a budget, for a gift that fits a description, or for the best running shoe for flat feet, and the assistant answers with specific products rather than a list of blue links. The moment of choice now happens inside the conversation, before the shopper ever opens a website.

That shift breaks the assumption every retail stack was built on: that a human will land on a product page, read it, and click buy. When an agent is the reader, the catalog, prices and availability have to be expressed in a form a machine can parse and trust. A beautiful product page the agent cannot read is, for agentic purposes, invisible.

This is the problem an Agentic Commerce Platform solves, and it is the category SkuLift coined. Rather than treating each assistant as a one-off integration, the retailer maintains a single canonical catalog and the platform makes it discoverable and buyable everywhere an agent shops on the customer’s behalf.

For a retailer, being agent-ready means three things hold true at once. The catalog is machine-readable so an agent can match an item to a shopper’s intent. Pricing and availability are live and accurate so the agent quotes an offer the store will honor. And a checkout path exists over the protocol the assistant speaks, so the agent can complete the order rather than hand the shopper off.

Because the major assistants speak different protocols, retail presence is plural by nature. ChatGPT-mediated purchases run over ACP, Gemini authorizes payment over AP2, and Claude reads live context over MCP. A retailer present in one assistant but absent from the others has a partial shelf, and the agent ecosystem rewards consistent presence across every surface.

SkuLift maps the retailer’s one canonical catalog onto all three protocols, so the same inventory, prices and policies feed ChatGPT, Gemini and Claude alike. The shopper sees one coherent brand whichever assistant they happen to be using, instead of three drifting versions of the same store.

SkuLift coined the Agentic Commerce Platform category, and retail is its most direct application. The retailer integrates one canonical catalog and price book; the platform expresses it as ACP for ChatGPT, AP2 for Gemini payments and MCP for Claude context, so an agent can find, validate and buy a product wherever the shopper is conversing.

Because everything is driven from a single source of truth, prices, stock and promotions stay aligned across assistants. A markdown applied once propagates everywhere; an out-of-stock size disappears everywhere. The retailer avoids the contradiction of an agent quoting a price or availability that the store will not honor.

The platform also measures whether the retailer is actually being recommended and transacted against relative to competitors, sampled from real agent answers. That share of voice tells a merchandising team exactly where its agentic shelf is strong and where the catalog or pricing data needs work to win more recommendations.

There is no fixed price list for agentic retail because the work scales with the catalog and the protocol coverage a retailer needs. A focused brand with a few hundred SKUs and a single assistant is a different engagement from a marketplace-scale catalog buyable across all three. Pricing is therefore quoted on request, scoped to the retailer’s catalog, protocol coverage and measurement needs.

Value is measured the way a retailer already thinks about a channel: presence, recommendation share and completed transactions. The platform reports how often the retailer is cited and bought against competitors inside assistants, so the investment is judged on agentic shelf share rather than on a feature checklist.