Agentic Commerce
14+
150+
$1B+
What agentic commerce is, and why your store is already in it
Agentic commerce is commerce where an AI agent acts for the buyer: searching, comparing, filling a cart and completing the purchase, often without the buyer ever opening your store. For a merchant it means the storefront now serves two customers at once. The person, and the software acting on their behalf.
This is not a forecast you have time to think about. If you sell on Shopify, your store already has an MCP endpoint and your products are already syndicated into AI shopping surfaces through the Shopify Catalog. Nobody asked you to opt in. So the question was never whether to participate. It's whether the version of you that agents are reading is any good.
Most stores fail agents silently. Products described only in prose, prices buried in scripts, availability nowhere machine-readable, checkout hostile to automation. The agent doesn't complain. It just buys elsewhere.
What we design and build
Agent-readable catalog
Products as structured facts, not just photography and prose. Sizes, materials, compatibility, availability, price: parseable at the source. That's agent-ready product data, and it's the foundation everything else stands on.
Queryable discovery
Agents don't scroll. They query. Product discovery built on real attribute data serves precise answers to precise questions, for agents and for the humans using your search bar.
Transactable checkout
The last metre matters. Checkout customisation that supports automated flows instead of breaking them, without weakening fraud posture or the human path to purchase.
Storefront architecture
Agent-readiness holds up best on a composable stack. Hydrogen with Sanity and Algolia gives structured data a native home; it's the platform design we reach for when catalogs are complex.
Shopify or Die
Insights into the current and future state of Shopify Plus commerce. Headless architecture, agentic commerce, integration strategy, and the engineering decisions behind stores that scale.
Start with a score, not a pitch
Every engagement starts with a score, not a pitch. The Agentic Commerce Assessment measures how your store performs for humans, search engines, and AI agents, and hands you a prioritised backlog either way.
From there it's design and engineering together: storefront UI and UX decided on data, product data modelled properly, and the whole thing operated as commerce infrastructure rather than a launch-and-leave project.
What actually changes in your store
The protocol layer stopped being theoretical. The Universal Commerce Protocol is an open standard Shopify co-developed with Google, backed by Amazon, Meta, Microsoft, Salesforce, Stripe, Etsy, Target and Wayfair. It covers the whole journey: discovery, cart, checkout. MCP servers are the implementations sitting under it, and every Shopify store has one. Yours included.
So the interesting question isn't whether agents can reach your store. They can, today. It's what they find when they get there.
Your catalog is already syndicated
Shopify Catalog standardises and enriches product data, then feeds it to AI surfaces like ChatGPT, Copilot and the Shop app. This happens by default. Shopify's own number: AI searches powered by its Catalog convert at twice the rate of ones running on scraped data. Read that backwards, which is the way that should bother you. If agents are scraping you instead of reading clean structured data, you're converting at half.
Structure beats copywriting
An agent comparing three products reads attributes, not adjectives. Size, material, compatibility, stock, delivered price. It reads your policies too, because it will check your returns terms before it recommends you. That's agent-ready product data, and at the discovery stage it's the whole game.
Test it yourself, in five minutes
Open an AI assistant. Ask it to find a specific product on your store, check whether your size is in stock, and tell you the delivered price. Whatever it gets wrong, or confidently invents, is your backlog. Most brands are surprised. Not pleasantly.
The part nobody is selling you
Everyone is optimising for discovery. Almost nobody has thought about what happens next: an order from a customer who never saw your site, never opened your welcome email, and will ask their agent, not your support team, to handle the return. Winning the order and losing the customer is the real risk here. We build for the whole lifecycle, which is why this work runs into systems and integrations instead of stopping at the product page.
Frequently Asked Questions
What is agentic commerce?
Commerce where AI agents act on behalf of buyers: researching, comparing, and completing purchases. For merchants, it means building storefronts that agents can read, evaluate and transact with, alongside the humans they already serve.
Do we need to opt in for our store to appear in AI shopping?
No, and that's the part most brands miss. If you sell on Shopify, your store already has an MCP endpoint and your products are already eligible to appear in AI shopping surfaces through the Shopify Catalog. You're in it by default. The work isn't getting in. It's making sure the data representing you is accurate enough to win the comparison.
What is the Universal Commerce Protocol (UCP)?
An open standard Shopify co-developed with Google that defines how AI agents transact with merchants, covering the full journey from discovery through cart to checkout. It's backed by Amazon, Meta, Microsoft, Salesforce, Stripe, Etsy, Target and Wayfair. MCP servers are the implementations that sit underneath it, and every Shopify store has one.
How do we test whether our store works for AI agents?
Open an AI assistant and ask it to find a specific product on your store, check whether your size is in stock, and tell you the delivered price. Whatever it gets wrong, or invents, is your backlog. It takes five minutes and it's usually uncomfortable.
Are AI agents actually buying yet?
Early, but moving fast, and the research phase is already agent-heavy. Being ready early is cheap: it's mostly structure and data discipline. Retrofitting under pressure later is not. The brands agents can already parse will be the ones agents already trust when volume arrives.
What makes a store agent-ready?
Structured product data agents can parse, discovery surfaces they can query, clear pricing and availability, and a path to transact. In practice: agent-ready product data, product discovery built on real attributes, and checkout that doesn't fight automation.
Does designing for agents hurt the human experience?
No, and this is the part most people miss. The foundations are identical: structure, speed, clarity, honest information. Stores built well for agents are faster and clearer for people too. We design both surfaces together, which is why this sits in our design practice, not a technical afterthought.