Opus Research has been tracking conversational commerce since Dan Miller coined the term in 2011. For more than a decade, the promise remained the same: consumers would use natural language to shop, and systems would interpret their intent. The reality was more modest. Chatbots answered questions. Some took orders. Most created friction. 

That changed in 2025. Large language models made agents capable enough to hold context, reason through goals, and complete multi-step tasks. Retailers began building AI shopping companions. Walmart introduced Sparky. Williams-Sonoma announced a culinary companion that blends product discovery with lifestyle advice. Each represents a bet that brands can deepen loyalty by meeting customers inside an intelligent, goal-oriented interface. 

But the bigger shift happened at the platform level. In October, OpenAI turned ChatGPT into a native checkout surface with “Buy it in ChatGPT” and open-sourced the Agentic Commerce Protocol (ACP) so any merchant could plug in. Now Google has responded. At NRF 2026, Sundar Pichai introduced the Universal Commerce Protocol (UCP), co-developed with Shopify, Etsy, Wayfair, Target, and Walmart, and endorsed by more than 20 additional partners including Visa, Mastercard, Stripe, and American Express. 

Agentic shopping has arrived. And two of the most powerful technology companies on earth are racing to define how it works. 

What UCP Actually Enables 

UCP is Google’s attempt to turn commerce into something an AI agent can reliably execute, not just discuss. Instead of every retailer building one-off integrations for every assistant or marketplace, UCP standardizes how an agent discovers what a merchant can do and then carries out actions across the full commerce lifecycle. 

The protocol supports checkout, identity linking (for loyalty and membership), and order management including status checks, returns, and cancellations. Google offers two integration paths: a native checkout via APIs, and an embedded checkout option for merchants with complex flows that need customization. The embedded fallback is an acknowledgment that real-world commerce edge cases will not normalize overnight. 

UCP will first power a new checkout feature in AI Mode in Search and the Gemini app for eligible U.S. retailers, with Google Pay and saved Wallet credentials handling payment. Merchants remain seller of record. Customer relationships stay with the brand. 

How UCP Compares to ACP 

Both protocols aim to make the AI agent the place where commerce resolves while assuring merchants they keep control. The differences are scope and starting position. 

UCP is broader. It covers end-to-end commerce wiring, including post-purchase workflows like tracking, returns, and support. It builds on Google’s existing commerce infrastructure, consisting of Merchant Center feeds, structured product data, and years of integration with retailers at scale. Google also designed UCP to be interoperable with its Agent Payments Protocol (AP2), Agent2Agent (A2A), and Model Context Protocol (MCP). 

OpenAI’s ACP is more purchase-centric. It is tightly tied to making checkout happen natively inside ChatGPT, with delegated payment support through Stripe’s Shared Payment Token as its first implementation. OpenAI’s advantage is the conversational surface itself. ChatGPT already knows what customers want because they are asking it questions all day. 

Both are distribution plays. Google anchors UCP in Search and Gemini. OpenAI anchored ACP in ChatGPT. The real contest is over which protocol attracts the most merchants and which conversational surface consumers trust for transactions. 

Why CX Leaders Should Pay Attention 

Once an agent can transact, it can also initiate the interactions that generate a large share of customer contacts: order status inquiries, cancellations, returns, refunds, and exception handling. UCP is explicitly designed to support those post-purchase workflows. That shifts pressure onto CX-owned assets. 

Policy becomes executable. Knowledge needs to map to actions. And the customer experience of your brand increasingly happens inside someone else’s interface, paced and phrased by their agent. 

This is not a distant scenario. Google’s protocol is live now, with Shopify already enabling its merchants to sell through AI Mode and Gemini. The merchants who wait risk being invisible to the agents their customers are learning to use. 

What to Do Next 

Treat agentic commerce as an operating model shift, not a channel experiment. 

Start by ensuring your order, inventory, and policy data are clean and machine-accessible. Decide where automation should stop and a human should step in. Instrument the full journey so you can see where agent-driven flows create friction and drive contacts back to your support team. 

Look for an upcoming Opus Research report on agentic commerce, with a deeper dive on UCP vs ACP economics, control points, and the practical integration and CX implications for retailers and platforms.

The winners in this next phase will be the retailers who show up where the agent lives, keep their data honest, and deliver when the agent says “buy.” 

Categories: Conversational Intelligence, Intelligent Assistants, Articles