Shopify just made the case for us
The catalog is now the interface
A couple of weeks ago I wrote about agentic commerce starting in the wrong place. That the brands likely to do this well are not the ones with the boldest AI roadmaps, but the ones whose operating models can actually move when a decision is made. That the visible work, strategy and tooling, sits downstream of the unglamorous work, product data and operating model alignment.
Then Shopify Editions Spring 26 landed this week, and a good chunk of it reads like the same argument with a product roadmap attached.
What they actually shipped
The headline of the agentic section is not “your brand in AI.” It is “your products optimised for AI.” That distinction matters. Shopify has decided, at a platform level, that the unit that matters in agentic commerce is the product, not the brand.
The clearest piece is Shopify Catalog. It automatically standardises and enriches product data so AI models can reason about it, and Shopify claims data it syndicates drives twice the conversion in AI chats. Around it sits the rest: checkout inside Copilot via the Universal Commerce Protocol, an agentic plan that lets non-Shopify brands sync products into Catalog, and a Catalog API that lets agents pull real-time pricing, availability and variants for products at request.
The product catalog is now the interface between your business and every AI agent that might recommend you.
Why this is good news, and a trap
It is good news because it confirms the thing we have been saying. Product-level visibility depends on structured, enriched, machine-readable product data. Shopify has now built tooling for exactly that and put a conversion number next to it. The 30% to 40% of the problem that Glara put in the “enrich the catalog” bucket at Pulse now has a button.
The trap is assuming the button is the job.
Catalog enriches what you feed it. It does not fix what you feed it with. If your product data is inconsistent across your PIM, your storefront, your marketplaces and your 3PL, then automatic enrichment makes that inconsistency travel further and faster. You do not get clean data in AI channels. You get the same mess, syndicated, with a higher conversion rate on the wrong information.
That is the 60% to 70% nobody is selling a tool for. It is not catalog enrichment. It is whether the brand can describe its own products the same way everywhere they live, and whether the operating model can keep them that way as the business changes.
The bit Editions does not solve
Editions makes the front of the stack easier. It does almost nothing for the back.
It does not tell you which system is the source of truth for a product attribute when three systems disagree. It does not resolve who owns product data when ecommerce, merchandising and operations all touch it, and none of them is accountable for it. It does not align your head of ecommerce and your IT director on a shared priority list, which in our experience is the single thing that determines whether any of this moves.
A platform can make your products legible to agents. It cannot make your organisation coherent enough to keep them legible. That work is internal, slow, and has no vendor category. It is also the work that decides whether the tooling pays off.
What we would actually do about it
Before you spend a quarter on agentic strategy, here are three unglamorous checks to carry out.
One. Pick your source of truth for product data and write down which system wins when systems disagree. If you cannot name it, that is the project.
Two. Name who is accountable for product data quality across the business, not who manages the PIM. These are different people more often than they should be.
Three. Get ecommerce and IT into one priority list for this work. Not aligned in principle. The same list, in the same order.
Do those three and Catalog enrichment becomes worth doing, because you are enriching something coherent. Skip them and you are buying reach for data you have not agreed on.
The brands that win the agentic shift will not be the ones who adopted Catalog first. They will be the ones who had something worth syndicating when they did.