Skip to content

Where should a brand start with agentic commerce?

Jeremie Werner

Founder of Conversion Club · ikko

View author

Short answer

Start with catalog quality and observable buyer journeys. Check market and channel eligibility before investing in AI checkout, then review three months of evidence.

The answer in detail

Start with discovery, then measure the route to purchase. Agentic commerce means an AI assistant searches, compares and sometimes purchases on a shopper's behalf. A readable catalog, reliable referrals and three months of observation are usually more useful starting investments than a speculative payment integration. Choose checkout work according to channels and buyers you can actually serve.

The benchmark

An assistant can affect a purchase at several points: finding suitable products, comparing their attributes and policies, referring the shopper to a store, or initiating a supported checkout. These are different behaviors. A recommendation appearing in a conversation is not proof of a completed delegated purchase, and a referral is not a measure of everyone who saw the answer.

Shopify’s English merchant guidance provides a practical operational starting point: identify which catalogs and channels expose your products, determine where payment takes place, and reconcile resulting orders with the route used. Those observable store facts are more useful for an initial plan than applying another market’s adoption percentage to your business.

Availability is a separate question from adoption. An independent October 29, 2026 check of Shopify's English documentation distinguishes ChatGPT referrals to merchant checkout from direct checkout on other activated channels. That replaces the source's France-centered available/unavailable table, whose launches and regional exclusions should not become worldwide operating instructions.

Question What to establish for your store
Can an assistant find the product? Catalog inclusion and actual recommendations
Where does the buyer pay? Merchant checkout or the specific channel's direct checkout
Can this customer use it? Buyer location, product, channel and payment eligibility
What does the merchant control? Catalog, settings, policies and fulfillment

Shopify's eligibility guidance is conditional; inclusion does not guarantee placement. Recheck it for the intended market before commissioning integration work.

The rule

Invest where shoppers make decisions today in your markets. For many stores, that means the assistant's answer before the click. Where an eligible direct-purchase channel exists, test it as a specific route with explicit costs and responsibilities; do not treat every assistant as a payment agent.

A staged plan remains useful: map exposure, measure journeys and risks over three months, then scale over six to twelve months using products and questions that demonstrably produce sales. These are planning horizons, not promises that a channel will launch by a given date.

The steps

  1. Map exposure and settings. Review Shopify's Agentic sales-channel section, what catalog data is shared, and which channels your store permits. The source describes automatic management being enabled by default; confirm the actual account configuration and terms. Its French-store case remains a regional example. Record where your products already appear.
  2. Repair catalog data. Start with EAN/GTIN, brand, attributes and Merchant Center. See why GTIN matters and catalog readiness. Price and availability must agree across feed, page and markup; an unavailable product recommended as in stock is a bad answer.
  3. Make facts readable. Put material, dimensions, shipping and returns in text. Keep structured data consistent, including shipping information where supported. A policy cannot be made accurate by markup alone.
  4. Measure recurring questions. Create an “AI assistants” GA4 channel using the measurement method. Monthly, repeat 10–15 buyer-style questions across relevant assistants: “Which [product] for [use], below [price], delivered to [market]?”; “Compare [brand] with [competitor] for [use]”; “Where can I buy [product] with free returns?” Record settings and recommendations. Read the visibility indicators over three months.
  5. Consider paid tests afterward. If a suitable advertising product is available to your business and audience, evaluate it separately. Verify actual account access, permitted markets and measurement before budgeting; an advertising announcement does not establish worldwide availability.
  6. Review quarterly. Check eligible channels and checkout routes again and update the roadmap. Record changes rather than relying on launch headlines.

The pitfall

Starting with payment infrastructure can consume the budget while barcode fields remain empty and competitors supply better product facts. Conversely, a France-only statement that agents cannot yet transact should not prevent a store in another eligible market from testing an available channel.

In an August 2026 Conversion Club interview, Payplug's Tomy Ferreira cautioned against predictions that ChatGPT would eliminate websites and expected no major revolution over the next six to eight months. Retain that as a dated expert view, not a current guarantee. The interview reinforces the case for sequencing investments around evidence.

What to tell your team

“We will align our catalog by December, measure assistant referrals for three months and review in April. Our current estimated assistant-influenced sales share is 2%; we will validate that estimate before expanding checkout work.” The figure and schedule are fictional; the checkout decision depends on our markets.

Sources

Further reading

The Playbook answers in short. Conversion Club articles take the time: methods, field cases and figures to go further.