Google's 60-Billion-Listing Bet on Measurable AI Commerce

Google's Shopping Graph now spans more than 60 billion product listings, and the company is betting that scale alone will not sell agentic commerce — measurement will. This week, Google paired with NielsenIQ (NIQ) to make AI-driven product recommendations auditable end-to-end, while Cordial opened its customer data platform to the AI assistants marketers already use. Together, the announcements answer a request enterprises keep repeating: connect trusted customer, product and interaction data to real-time decisions, without losing oversight of how those decisions get made.

The timing is not incidental. As customer data becomes an operational layer for AI agents, the stakes around data quality, shared definitions and governance rise sharply. An agent can only personalize responsibly if it understands what a customer signal means, can access the right context at the right time, and knows when to hand off to a human. Vendors that cannot demonstrate all three are increasingly being screened out of enterprise evaluations.

What does the Google–NIQ partnership actually cover?

The collaboration links Google's Shopping Graph — those 60 billion-plus listings — with NIQ's Commerce Intelligence stack. NIQ's Product Intelligence covers more than 250 million products, and its Store Intelligence tracks retail locations and sales across 90 countries. That reach matters for one specific workflow: letting an AI system determine whether a recommended product is relevant, available locally, and ultimately purchased.

Ashish Gupta, VP/GM of Merchant Shopping at Google, was candid about the dependency. "While every brand is pursuing AI-driven growth, autonomous agents are only as reliable as the data behind them," he said. "Part of our partnership with NIQ helps us examine how their rich catalog and availability data can best support new agentic commerce experiences."

Note the verb: "examine." The partnership, at this stage, is an investigation into whether NIQ's data can support agentic experiences — not a shipped, closed-loop product. The deal also includes NIQ integrating Google's open-source Meridian marketing-mix-model framework into its Cadence platform, which would let brands combine retail sales data with always-on measurement, experimentation and spend optimization.

For CX leaders, that promises closed-loop commerce measurement. The catch: someone still has to validate the recommendation logic and attribution models sitting between discovery and purchase. Neither vendor detailed methodology or sample sizes for those models, and that gap is where the promise currently outruns the proof.

Why does real-time correction matter for agents?

Tara DeZao, Senior Product Marketing Director at Pega, framed the operational requirement for CX teams: "If you get the interaction wrong the first time, it's really important to correct that interaction in real time. And that's where adaptive decisioning comes in."

Her point cuts to the governance question the week's announcements raise. Effective intelligence needs to support consistent decisions across products and services, maintain a clear audit trail, and offer a credible recovery path when automation fails. Two of those three — audit trail and recovery path — remain largely unsolved in agentic commerce, and none of this week's press materials specified how they will be delivered.

How does Cordial fit in?

Cordial launched a headless version of its embedded customer data platform, following an earlier headless AI infrastructure release this year. Through MCP architecture, enterprise customers can now pull real-time audience and customer data into Claude, ChatGPT, Microsoft Copilot and custom agents, and plan programs directly inside those assistants.

The move tracks an API-first trend with a practical rationale: marketers work where they already work. Whether real-time CDP data inside a chat assistant produces better programs than a traditional UI is a claim Cordial has not yet substantiated with published benchmarks.

The direction, though, is consistent across all three vendors: as AI assistants take on larger roles in discovery and buying, brands will demand visibility beyond clicks and recommendations — and vendors that cannot show whether an AI-led journey produced a meaningful customer outcome will struggle to keep enterprise accounts.