Your next replatform is an agentic commerce project

A robotic arm placing a parcel into a shopping basket

By Bret Williams, Founder/CEO, novusweb

AI shopping agents read your product data, not your storefront. If you're planning a move to Shopify Plus, the migration is the best chance you'll get to rebuild that data so agents can find, compare and sell your products. Scope it in from discovery, because retrofitting it after launch costs more and usually doesn't happen.

What changed in the last 14 months

Two announcements moved agentic commerce from demos to checkout:

  • On September 29, 2025, OpenAI and Stripe launched Instant Checkout in ChatGPT along with the Agentic Commerce Protocol, an open standard for agents and merchants to complete a sale.

  • On January 11, 2026, Shopify and Google introduced the Universal Commerce Protocol (UCP), another open standard for agent transactions, alongside Shopify's Agentic Storefronts for syndicating product data to ChatGPT, Microsoft Copilot, Google AI Mode and Gemini.

The protocols will keep evolving. What won't change is what every agent needs from a merchant: accurate, structured, current product data, and a checkout it can complete.

Two things an agent does with your store

It helps to separate them, because they need different work.

Discovery. An agent decides whether to recommend you. It leans on product attributes it can parse (materials, dimensions, compatibility, price, availability), on consistent information about your brand across the web, and on content that answers real buying questions. You don't opt in to discovery. Your data is either good enough to be picked or it isn't.

Native selling. The shopper completes the purchase inside the AI conversation. That requires the platform connection (on Shopify, the Catalog and Agentic Storefronts), plus inventory and pricing that are right at the moment the agent asks.

Why the migration is the moment

On Adobe Commerce (Magento) and Salesforce Commerce Cloud, product information tends to sprawl. Years of custom attributes, attribute sets that mean different things in different categories, specs pasted into description HTML, and variants modeled three different ways. Humans browsing a product page can work around that. An agent comparing your product with 40 others can't.

A migration already forces you to map every product field from the old platform to the new one. Cleaning the data during that mapping costs a fraction of a separate project later. Here's what I'd put in scope:

  1. Titles an agent can read. Brand, product type and the attributes that distinguish it, in literal language. Save the clever names for the page copy.

  2. Specs in structured fields. Move materials, dimensions, weight, care, compatibility and certifications out of description HTML and into typed metafields. Map them to Shopify's standard product taxonomy.

  3. One way to model variants. True variants under one parent, each with its own image, price, inventory and GTIN.

  4. Identifiers on everything. Valid GTINs (UPC or EAN), consistent SKUs and manufacturer part numbers. Agents use them to match your product to what the shopper asked for.

  5. Structured data on the storefront. Product and Offer schema in JSON-LD that updates with variant, price and stock changes.

  6. Inventory and pricing that are right in real time. This is where ERP and warehouse integrations matter. If stock syncs every few hours, an agent will sell items you don't have.

  7. Market-level pricing. Fixed local prices per market, so an agent quotes the same price the checkout charges.

What the platform does and what you still own

Shopify handles much of the protocol work, including Catalog syndication and UCP support. It can't fix your data or your integrations. On a 2015 project I led, moving more than 70 custom stores to Shopify Plus, the work that decided success wasn't the storefronts. It was the ERP backend I architected behind them, which let the first flash sale on the new setup take more than $800K in its first hour without a hiccup. Agentic selling raises the stakes on that same backend: inventory, pricing and fulfillment now have to be right for shoppers who never see your site.

Questions to ask before you scope a migration

  • Which of our product attributes live only in descriptions or images?

  • How many ways do we model variants today?

  • What share of our SKUs have valid GTINs?

  • How often does inventory sync from our ERP or 3PL, and what happens when it's wrong?

  • When someone asks ChatGPT for the best product in our category, do we show up, and is what it says about us correct?

If you don't know the answers, that's what a discovery is for.

About the author

Bret Williams wrote three books on Magento for Packt. He has led migrations from Adobe Commerce (Magento), Salesforce Commerce Cloud and WooCommerce to Shopify, and writes about getting stores ready for AI shopping agents. novusweb advises merchants on replatforming to Shopify Plus and stays on their side through the project. Book a free intro call.

novusweb®

Full, 360 degree e-commerce agency with over 20 years experience building innovative online solutions.

https://www.novusweb.com
Previous
Previous

How long does a Shopify Plus migration take?

Next
Next

Adobe Commerce 2.4.6 is out of regular support. What now?