Agentic commerce demands better data - most retailers aren’t prepared

For years, digital commerce strategies were built around simple assumptions: humans search, humans browse, and humans decide. These assumptions no longer hold true. 

Today, AI agents are increasingly acting on behalf of consumers — discovering products, comparing options, and in some cases completing transactions. It’s become clear that large language models (LLMs) have ascended to a legitimate shopping channel, as many major global retailers have shifted their strategy to incorporate agentic commerce. 

The momentum is real. The preparation is not. On a scale of one to 10, the average retailer is only a three or four when it comes to preparedness for agentic commerce. The gap has nothing to do with conversations in boardrooms. It is about data.

Discovery has moved upstream

Retailers are already seeing the signals. Bot traffic has surpassed human traffic across large portions of the web, and as organic search declines, discovery is shifting more toward answer engines and AI assistants.

A growing number of shoppers now begin their journey by asking an LLM for advice rather than typing into a search box. Their prompts are longer, richer, and far more specific than traditional searches, and they convert at surprisingly high rates.

What makes this shift so significant is where decisions are now being shaped. Awareness is no longer formed primarily on search engines and brand-owned channels. It is formed earlier, inside systems that rely entirely on data to determine what appears and what does not.

Why most retailers are invisible to agents

AI agents do not shop the way people do. They do not rely on emotion, become enticed by marketing copy, or infer meaning from design. They operate on context, structure, and clarity.

To recommend a product, an AI agent needs to understand exactly what that product is, who it is for, whether it is available, and why it is relevant. That requires data that is explicit, complete, and accessible. Most retail websites were never designed for this.

Product data is often fragmented across systems, leading to attributes that are incomplete or inconsistent. Inventory and fulfillment information is often delayed or siloed. Pricing and promotional offers are disconnected from availability. Even when the data exists, it is rarely structured in a way that machines can reliably interpret.

In response, agents hesitate, and when agents hesitate, they move on. Products that cannot be confidently understood are effectively removed from the agent-driven funnel.

The new battleground for awareness

This is why agentic commerce is fundamentally a data challenge, not a channel challenge.

It is inherently impossible to advertise to AI agents, so brands will win by building data foundations that allow agents to act with certainty on behalf of consumers. In the agentic era, the retailers with the strongest data will capture awareness. 

This data must serve two worlds at once. The human web still matters, but the agentic web has different requirements: denser metadata, clearer structure, and richer context. Retailers that harmonize these needs will be positioned to compete. Those that do not will be invisible.

Three priorities for agent-ready retail

While every retailer’s journey is different, there are three priorities that consistently matter most.

• First, make product data explicit and machine-readable.
  This means moving beyond basic descriptions. Most retailers already have product information, but it too often is incomplete, inconsistent, or optimized for human consumption. AI agents need product data that is explicit and unambiguous, meaning it contains attributes like size, dimensions, compatibility, materials, and variants. If an agent cannot confidently compare options, it will not recommend them.
• Second, provide meaningful, contextual semantic summaries.
  Semantic summaries help explain who a product is for, what problem it solves, and why it is relevant in a given moment far more effectively than structured data alone. Two products may share similar specifications, but additional semantic context can help agents understand how they serve very different audiences or use scenarios. These summaries should be written with both humans and machines in mind, enabling more relevant discovery and more personalized outcomes.
• Third, organize products by the problems they solve, not just categories.
  As discovery becomes intent-driven, shoppers are more likely to describe the outcome they want to achieve than ask for a specific item name. These tags give AI agents a way to match products to intent-rich prompts, even when brand names or product types are never mentioned. Retailers that tag and structure products by the problems they solve are far more likely to surface in agent-led discovery.

When incorporated, these steps ensure brands are prepared to reap the rewards from the trust shoppers have in AI outputs. Visibility into data processes becomes clearer, and there’s greater operational alignment that allows AI-driven discovery to translate into real business action.

Consumers reward brands that are relevant, responsive, and easy to buy from – wherever discovery happens. That requires connecting what happens at discovery to planning, forecasting, and inventory decisions. When retailers can see what agents are recommending and how consumers are responding, they can plan better, personalize more effectively, and deliver experiences that earn repeat engagement.

The window is closing

The data is clear. Discovery patterns have changed, traffic flows have shifted, and conversion behavior is evolving.

Agentic commerce is already here, and retailers now face a narrow window to act. Those that invest in strong, harmonized data foundations will be visible where decisions are being made. Those that delay will find themselves excluded from the very systems shaping the future of shopping.

In the agentic era, data readiness is not optional. It is the price of participation.

Balaji Balasubramanian is president and chief product officer of SAP CX.