Google's Universal Commerce Protocol (UCP) recently launched at National Retail Federation (NRF) conference with a roster that's hard to ignore. It has Shopify, Walmart, Target, Etsy, and Wayfair as co-developers, plus endorsements from Visa, Mastercard, Stripe, PayPal and 20+ others.
It’s the new blueprint that powers agentic experiences across the commerce ecosystem. But is it the one and only choice? Let’s look at what brands should actually consider when looking to implement AI agent-based shopping.
UCP benefits the entire commerce ecosystem
UCP creates a standardized way for AI agents to interact with commerce systems, from product discovery right through to checkout and order status. Instead of building one-off integrations for every AI assistant, retailers can make their capabilities available once and be discoverable across Google's AI surfaces (Search AI Mode, Gemini). Over time, they will be usable by any agent that supports the protocol.
Google already has Merchant Center feeds for millions of businesses. UCP now turns that existing data into an agentic commerce layer.
Meanwhile, OpenAI's Agentic Commerce Protocol (ACP) has been live since September, powering ChatGPT Instant Checkout. ACP takes a different approach than UCP. It’s tightly scoped around the purchase moment, built for fast integration through Stripe.
What UCP will standardize right now
Despite the big vision ("end-to-end lifecycle"), the initial launch focuses on three concrete capabilities:
- Checkout
- Identity linking (enables loyalty/personalization)
- Order tracking
Identity linking is not magic. It's OAuth 2.0 authorization code flow, packaged in a consistent protocol shape that agents/surfaces can adopt broadly.
UCP is designed so that the merchant remains the merchant of record, which matters for control, liability, and existing operating models. At the vision level, UCP is vendor agnostic, however its initial release relies on the Google platform to facilitate product search and catalog feed ingestion.
The practical reality for brands
So between UCP and ACP, could you pick just one? No, you’ll likely need to support both UCP and ACP, just like you build for iOS and Android.
But protocol support is table stakes. The harder question is whether your product data, checkout logic, and post-purchase operations are actually ready for agent-driven commerce. An AI agent can't represent your loyalty program if it's not machine-readable. It can't apply your business rules if they only exist in checkout page code.
UCP doesn't remove complexity, it redistributes it:
- The ecosystem is coming together. Major retailers, payment networks and platforms are aligning around these standards, which signals real commitment.
- The integration landscape just got more complex. UCP, ACP, A2A, MCP, AP2...The acronym soup is real, and each has a role.
- Your data infrastructure matters more than ever. Structured product feeds, clean taxonomy and exposed business logic become competitive advantages when agents are your intermediaries.
- Customer relationships don't disappear, but they change. UCP explicitly keeps retailers as merchant of record. You still own the relationship, but how customers experience your brand increasingly happens inside someone else's interface.
While the protocol is designed to be highly flexible, different businesses will need different levels of customization to support checkout. Most of that complexity comes from two main areas:
- Product complexity: Simple, well-defined Stock Keeping Unit (SKUs), without complex configuration options, are the best fit. Things like bundles and customizable product /services will require additional engineering.
- Checkout/selection complexity: Thorough product comparison, complex stock checks and multi-step checkouts will add an additional layer of development.
This means that CPG and retail merchants will benefit from the more vanilla UCP implementation, while other domains (e.g. event ticketing, automotive, services, insurance) will require additional engineering force to benefit from agentic commerce.
A practical next step
When it comes to making a choice, your question shouldn’t be "should we bet everything on agentic commerce?", but rather ask yourself which one or two journeys are worth making agent-ready first.
Here are some common starting points:
- Product discovery based on relevant data
- Authenticated checkout for a subset of catalog
- Loyalty-aware offers
- Order status and fulfillment tracking
Working with Empathy Lab
We're not here to tell you agentic commerce will transform everything overnight. We're here to help you figure out what's actually worth doing now, and make sure your investment delivers. Together, we’ll work through these steps:
- Assessment: What does your current architecture expose to agents? Where are the gaps in identity, checkout and data contracts?
- Prioritization: Which protocols matter for your customer base? Where should you invest first or rather wait?
- Pilot with measurable ROI: One journey, clear KPIs, real business outcomes. We go beyond a tech proof-of-concept.
The brands that do well here won't be the ones who bought the hype earliest. They'll be the ones who understood the technical reality, built accordingly, and measured what matters.
If you're trying to separate signal from noise and pick the right first experiment, get in touch with our commerce experts here.
Contributors in this article