UCP just made product data a buying surface
The unsexy follow up to Monday's article
If you read Monday’s piece on UCP, this is the unsexy follow-on. If you missed it, it’s here www.commercethinking.com/p/ucp-ai-and-agentic-commerce.
That was the why. This is the what now.
Context
Google’s Universal Commerce Protocol is an attempt to make agent-led buying normal. Shopify’s been in the room, so commerce systems can talk to agents without every brand doing one-off work for every assistant.
Shopify’s write-up is here www.shopify.com/uk/ucp.
The thesis is simple. Your store is about to get a new kind of customer, and it doesn’t browse. It queries. Back in June 2025, we said it plainly. Your product data isn’t just powering your website or your ERP anymore. It’s becoming your interface. And not just for your customers, for the AIs they’ll increasingly use to shop, research, and transact on their behalf.
More bluntly. In the near future, your product data will be your API for the world.
UCP makes that literal. It’s the agentic web idea moving from theory to plumbing, where agents ask structured questions and expect structured answers.
And it doesn’t let you hide.
What changes in the back office
Most teams will treat UCP like a channel launch and a tech project. It’s both, but the order matters.
If you can’t describe what you sell, what it costs, and how it gets to the customer in a way a machine can trust, you won’t get the sale.
UCP isn’t just product discovery. It’s checkout, discounts, and loyalty logic getting pulled into a conversation.
If that stuff is fuzzy internally, the agent will find the edge cases faster than your best CX lead.
That’s why we keep banging on about product ops and semantic clarity, not pretty PDPs, because machines don’t buy the vibe.
Do the product ops, properly
Start with your top ten sellers and try to answer the buying questions using only structured fields. No free text descriptions, no Slack messages to the merch team, no someone in CX who just knows.
If you can’t do it internally, you definitely can’t do it when an external agent is asking.
Three bits of homework to get you UCP-ready.
Decide where the truth lives for product types, variants, and attributes, and stick to it.
Turn your operational promises into data. Delivery windows, shipping rules, return eligibility.
Treat agent orders like a real channel. Tag them, report on them, and make someone accountable for the inputs that drive them.
Don’t get hung up on whether you need a PIM, a new feed tool, or a big rebuild. You need the basics nailed, consistent product types and variants, and clean normalised attributes.
Then make it legible for machines. Make your category hierarchy explicit. Bake Schema.org and JSON-LD markup into your site architecture.
After that, put governance in place so it doesn’t degrade over time.
Because the point still holds. The gap isn’t closed with more filters or better CMS copy; it’s closed by turning your product data into a structured, queryable, machine-friendly asset.
The big reveal
UCP is going to drag a lot of hidden complexity into the light. If your promise differs by region or basket value, that’s fine, but it has to be represented as structured logic, not folklore.
The agent will ask boring questions with brutal certainty. Is this size in stock? What’s the delivery promise to that postcode? Is this item final sale?
The other shift is that you’ll get orders that didn’t start on your site, but they still hit your OMS, your finance recon, and your returns flow. Set up the mechanics now so you can tell promo-driven margin movement from channel-mix margin movement.
If you’re on Shopify, a lot of the plumbing’s handled already. That doesn’t change the homework, it just reduces the integration tax.
If you’re not on Shopify, you’ve got a decision. You either use a connector that’s already done the work. Shopify is openly offering itself for that role. Or you wait for your platform to catch up.
Waiting might be rational, but do it with your eyes open. This isn’t an R&D conversation anymore.
We wrote that your data is either a bridge or a wall. UCP is going to make it obvious which one you’ve built.
Closing point
Clean the catalogue, make your operational rules explicit, and tag and track the channel. That’s how you win.
And don’t ignore the ongoing bit. Someone has to wake up in the morning owning the data.
Useful links
The following links are practical guides to sorting out your underlying product data, it’s the stuff your Merch and commercial teams are desperate to get right - UCP is just another reason to nail it:
The original product-data piece is here:
www.commercethinking.com/p/your-product-data-isnt-just-for-humans
Bridging the gap between merchandising and tech to drive better decisions:
https://www.commercethinking.com/p/hey-tech-teams-heres-what-merch-needs
Striking the perfect balance in product data:
https://www.commercethinking.com/p/hierarchy-vs-attributes-debate