The Data Layer That Agentic Commerce Can’t Ignore

AI agents have made the move from search assistants to autonomous buyers. The biggest platforms are now racing to own that transaction layer, and Amazon’s latest move makes that concrete. Alexa for Shopping, built from the merger of Rufus and Alexa+, tracks prices and remembers context across devices. Then, with Auto-Buy, it completes a purchase the moment a product hits a target price, no cart review or manual checkout required.

Agentic shopping is a market projected to reach upwards of $5 trillion in global commerce volume by 2030. Amazon is hardly alone in trying to carve out a slice of that sector – Google, OpenAI and Perplexity have all moved aggressively into agentic shopping. That push is also increasing consumer comfort with AI participating in their purchasing decisions, with AI-driven traffic to U.S. retail sites surging 393% year-over-year in Q1 2026. 

The consumer experience is easy to understand. What’s less visible is the infrastructure that makes it possible. When an AI agent transacts autonomously, it trusts the product data completely. No human flags the wrong variant or the misclassified product. That makes the data layer the most consequential part of agentic commerce. 

The Foundation for Agentic Commerce

The conversation around agentic commerce is mostly about the agent — the model, the interface, or the experience. But the real determinant of success is not the agent itself, it’s the data layer supporting e-commerce. 

When a consumer asks an agent to reorder laundry detergent or buy the cheapest HDMI cable that can arrive tomorrow, the transaction appears simple. Behind the scenes, however, the agent must make a series of decisions. It needs to identify the exact product being requested, verify that product’s attributes and availability, compare options across multiple platforms, and connect the purchase to the right inventory and fulfillment network. 

Unlike a human shopper, an AI agent cannot rely on intuition to fill gaps or resolve inconsistencies. Every step in the transaction needs to be fueled by data that is structured and consistent across the systems involved. If any part of that foundation is missing or fragmented, the agent doesn’t pause to investigate. Instead, it works with what it has and transacts anyway, or fails to complete the transaction altogether, creating inconsistent and often unreliable user experiences. 

Where The Data Foundation Breaks Down

According to Mirakl, less than 1% of eCommerce product pages currently meet the minimum standards for recommendation by AI agents. Product identifiers vary between supplier systems and internal systems, while inventory status lags behind reality. Location data also lacks standardization across the fulfillment network. 

When humans are in the loop, bad data is manageable. A buyer notices the wrong variant, or a warehouse manager knows the inventory system is running behind. People fill in the gaps from experience; the problem is that an AI agent has no such instinct. It reads the data it’s given and completes the transaction, or fails to complete it altogether.

When the data is wrong, the transaction is wrong. A completed purchase of the wrong product, routed to the wrong location, at the wrong price. The agent performed exactly as designed, it was the data underneath that didn’t. 

The Identity Layer Every Agent Depends On

Poor data quality costs organizations at least $12.9 million a year on average. That figure comes from environments where humans are still in the loop. In agentic commerce, that number only grows. Every automated transaction the agent gets wrong compounds the cost. 

A pair of running shoes listed as “trail runners” on one platform and “all-terrain sneakers” on another is the same product, but an AI agent can’t confirm that without a universal identifier. 

That’s what the product identifier embedded in a UPC barcode helps solve. These scannable signals can carry a GTIN (Global Trade Item Number), which are part of a global system of standards used by more than 2 million companies used to uniquely identify products in the global supply chain. Scannable GTINs give products a single, verified identity across every retailer, warehouse and platform that touches them. When an agent compares prices on those shoes across Amazon and Target the GTIN is what confirms it’s actually comparing the same product. 

Standards, like those from GS1 US, provide a common language for identifying products, locations, and entities, the same language AI agents need to communicate, coordinate and make accurate decisions. As more AI agents enter the supply chain — one managing procurement, another monitoring inventory, a third tracking logistics — they need to share data across systems built by different companies. Standardized identifiers make that possible. 

GTINs are part of that broader framework. For AI agents, the common language they provide is the difference between a transaction that completes accurately and one that does not.

Incomplete product data has the potential to remove a product from the agent’s consideration set entirely. For smaller brands, that’s an existential visibility problem.

Agent-Ready Data vs. What Most Companies Have

Getting to agent-ready data means ensuring every product in a catalog has a verified, globally unique identifier. It means attributes are complete and consistent across every system a trading partner touches. It also means inventory and pricing data are synchronized in real time, not updated in batches. Data built for human browsing doesn’t meet that bar.

Layering more AI on top of fragmented data produces faster, more confident wrong answers. The foundation has to be right first.

Gartner projects that more than 40% of agentic AI projects will be canceled by 2027, citing escalating costs, unclear ROI and inadequate risk controls. Data infrastructure quality is a direct input to all three. Companies building toward agentic commerce without auditing the data layer are solving the wrong problem.

Where Retailers and Brands Should Start

1. Audit product identifiers across every system. Every product in a catalog needs a verified, globally unique identifier. Missing or duplicated identifiers are the most common reason agents skip products entirely. 
2. Standardize attribute labeling. Size, weight, variant, and availability need to mean the same thing across every system a trading partner touches. Inconsistent labeling is invisible to humans and fatal to agents. 
3. Synchronize data across trading partners in real time. Most companies update inventory and pricing on a delay — hourly, nightly, or manually. An agent executing a purchase doesn’t wait for the next update. 
4. Treat data readiness as a prerequisite for any agentic AI investment. Companies building toward agentic commerce without auditing the data layer are solving the wrong problem. Remember that the agent is the last thing to optimize; the data foundation needs to be first.

The retailers and brands that invest in that foundation now will be more discoverable and transactable to the agents making purchasing decisions on consumers’ behalf. Those who wait will be competing for visibility in a system that was never built to find them. The agent is the visible part. The more important question is whether the data underneath it is ready.