How Universal Commerce Protocol will change commerce: Discovery first, transactions later

Google’s announcement of the Universal Commerce Protocol (UCP) is the clearest signal yet that agentic commerce has moved from concept to infrastructure. Co-developed with Shopify, Etsy, Wayfair, Target, and Walmart, and endorsed by Visa, Mastercard, Stripe, PayPal, and American Express, the UCP represents a coordinated industry commitment to a future where AI agents mediate the buying journey. 

Buying journeys are no longer anchored to websites and apps. Discovery, in particular, has already changed. Search results, conversational interfaces, voice interactions, and embedded touchpoints are converging into a single decision layer where machines decide what is relevant before a human ever engages.

Those machines do not experience brands the way people do. They do not respond to visual hierarchy, storytelling, or carefully designed journeys. They evaluate product information. Structured data, consistency, completeness, and confidence matter more than presentation.

This is a turning point in how product experience works. What once lived at the front end of the stack is moving into the operational core. But it is important to be precise about where the disruption is happening first, and where it will follow.

Discovery and transaction activities are not being disrupted at the same pace. 

Discovery has already shifted. Machines are actively mediating what products are surfaced, compared, and shortlisted. That change is visible today in search behaviour, retail media dynamics, and conversational interfaces. Brands are already winning or losing consideration based on how well their product data can be interpreted upstream.

Transactions are different. While infrastructure is being put in place, large-scale, agent-driven purchasing is constrained by trust, liability, payments, and operational complexity. Meaningful transaction volume through AI agents is unlikely to materialize overnight.

TLDR: Discovery-side disruption is immediate and irreversible. Transaction-side disruption is coming, but it will take time. Most commerce brands should expect real volume later in the decade, not next quarter.

For years, most commerce teams have been optimizing destinations: driving traffic, improving on-site conversion, and refining owned experiences. That model assumed the customer arrived, explored, and made a decision within an environment the brand controlled.

UCP accelerates a different assumption. Decision-making happens upstream, before a customer ever reaches your brand’s surface. AI agents assemble options, evaluate trade-offs, and narrow choices based on what they can reliably interpret. In many cases, they complete the transaction without a traditional visit at all.

When agents become the primary interface, your product data becomes the storefront. The quality of that data determines whether a product is even considered. This is not an abstract future state. It is already visible in how search, retail media, and conversational discovery are evolving.

In a protocol-driven ecosystem, visibility is not earned through clever UX or incremental media spend. It is a byproduct of readiness.

If data is incomplete, inconsistent, outdated, or difficult to validate, systems hesitate. And when systems hesitate, they move on. There is no partial credit for being almost ready. Products that cannot be confidently understood will fall out of consideration.

This is a difficult adjustment for organizations used to solving performance problems at the channel level. Under UCP, many of those problems originate much earlier, long before a campaign or experience comes into play.

The strategic play: Don’t bet on one protocol. Build the data and content layer that works across all of them: structured product information, semantic architecture, API-first infrastructure. Create your own optionality so you don’t get locked out of whichever surface wins.

Brands aren’t underperforming because they lack ideas or intent. They are exposed because their product information has grown organically, reactively, and in silos.

Data lives in too many places. Ownership is shared but rarely clear. Enrichment and approval still depend on manual effort. Local teams create exceptions that slowly become the rule. Over time, this creates an ecosystem that functions, but only with constant human intervention.

That approach was manageable when commerce was destination-led. But in a world where machines expect consistency and structure, it becomes a liability.

There is a tendency to frame AI-driven commerce as a race toward more advanced interfaces or smarter personalization. We think that misses the real constraint.

The organizations that will succeed are not those with the most experimental front ends, but those with the strongest foundations. Product experience management now sits directly on the path to revenue. But remember, Product Information Management (PIM) is a tool, not a strategy. The question isn’t “which PIM should I buy?” It’s “Is my product data structured, complete, and real-time enough to feed any of these integrations?”

Without that discipline, protocol adoption becomes fragile. Each new requirement adds manual work, risk, and cost. Over time, progress slows rather than accelerates.

In a UCP environment, the content supply chain plays a role similar to logistics or payments. It is not a creative function alone. It is an operational system that determines speed, resilience, and scalability.

When content operations are mature, organizations can adapt to new schemas, new markets, and discovery surfaces with confidence. When they are not, every change becomes a project, and every project becomes a bottleneck.

The difference is rarely ambition. It is whether content has been treated as infrastructure or overhead.

This shift is quieter than many AI announcements, but more consequential. It rewards teams that have done the unglamorous work of standardization, governance, and integration.

At DEPT®, we evaluate readiness across three domains:

• Content: Your site now serves two audiences: humans and AI agents. Content must be structured for emotional storytelling (humans) and machine retrieval (agents). At DEPT®, we believe systems should handle the friction so that customers can form the connection. The joy of discovery remains, even as AI removes some of the “work” from shopping. 
• Platform: Your infrastructure should outlast any single AI platform or protocol. Headless, API-first, composable architecture that can feed UCP, ACP, or whatever emerges next. 
• Data: It’s the semantic layer that makes content machine-readable and platform agent-ready. Schema implementation, entity optimization, real-time inventory and pricing, taxonomy that AI agents can parse without human interpretation.

Can machines understand your products well enough to recommend and transact them without human intervention?

If the answer is no, the gap is not creative or strategic. It is operational.

On a final note, we want to acknowledge the uncertainty of it all. Gartner predicts that over 40% of agentic AI projects will be canceled by the end of 2027. And even digital natives prefer shopping in-store, especially for luxury and beauty items. 

That being said, this vital infrastructure work pays off even if agentic commerce takes longer than predicted. It improves your owned experience, retail media performance, and search visibility. 

The downside of being ready early is zero. The downside of being ready late is existential.