Predictive Readiness: Why Product Data Is the Foundation of Agentic Commerce

Sitation has had a busy Q2 traveling to industry trade shows and conferences both to share our expertise and to gain new insight and vantage points for the latest evolution of e‑commerce. Like many years, there is a common thread, and like the last few, it begins with AI. Agentic commerce is not just the latest buzzword but rather an expansive frontier for how brands and retailers reach their ideal customers.

There is no shortage of conversation about what agentic commerce will look like. AI shopping assistants that book your travel, restock your pantry, or shortlist your next pair of running shoes. The conversation we want to start with, though, sits one layer below the shopping experience. Before any of that happens, an AI agent has to find your product, understand it, and trust it. That depends on something far less glamorous than a conversational interface: your product data.

What Agentic Commerce Actually Means

Gartner defines agentic commerce as a buying journey in which a buyer uses an AI platform or agent to discover, negotiate, decide, or transact. In other words, the agent does the work the shopper used to do. Gartner’s strategic planning assumption is that 20% of digital commerce transactions will be executed through AI platforms by 2030.

The change is not theoretical. OpenAI has reported roughly 2.5 billion ChatGPT prompts per day, with about 2.1% involving purchasable products, more than 50 million product‑related conversations every day on a single platform. Traffic to digital commerce sites originating from AI platforms was up 805% year over year on Black Friday 2025.

The Agent Doesn’t See your Homepage

For brands, this changes what good content means. The AI agent does not see your homepage, your hero video, or your influencer campaign. It sees a product feed, a structured data response, or a page it crawls and parses. Whatever it can interpret with confidence is what it will recommend. Whatever it cannot interpret, it skips.

Gartner puts it plainly: “Poor data structure will result in poor outcomes, regardless of product quality.” A brand can have a better product, a fairer price, and a stronger reputation, and still lose the sale to a competitor whose product data was simply more legible to the agent doing the recommending.

What “ready” Looks Like in 2026

Predictive readiness; being findable, preferable, and trustworthy before the shopper’s journey begins, comes down to three disciplines that sit squarely inside the product information management (PIM) world.

The first is structured attributes. Every product needs complete attribute coverage, organized with a consistent taxonomy and schema. Agents process information differently from humans. Where a person might forgive a missing dimension or guess at a material, an agent will filter the product out of the consideration set entirely.

The second is semantic, outcome‑focused content. Traditional product copy lists features. Agents need short‑form natural language that explains how a product is used, what problem it solves, and what kind of buyer it is right for. Specifications and use cases are not interchangeable, and brands need both.

The third is comprehensive evaluation data. Shoppers using agents can specify criteria they rarely articulate at the digital shelf, like country of origin, certifications, sustainability claims, ingredient or component details, returns, warranty terms and third‑party reviews. If your data does not meet those criteria, the agent will not surface your product to a shopper who asked for it.

Where the Data Needs to Go

Getting product data ready is only half the work. The other half is getting it to the right places. AI platforms are building direct ingestion paths. OpenAI’s Merchant feed, Google’s Universal Commerce Protocol, and Perplexity’s product integrations are all examples. Each one carries its own schema and requirements. At the same time, AI crawlers are reading public commerce sites, which means schema markup on product detail pages and the right permissions in your robots.txt file matter again, in a way they have not for years.

A newer layer is emerging in the background. The Model Context Protocol (MCP), introduced by Anthropic in late 2023, is becoming a standard way for AI agents to discover and access external data and tools. Leading PIM platforms are beginning to support MCP server creation, which would let agents query product information directly rather than waiting for a feed to refresh. This space is moving in weeks rather than quarters. Investments need to flex with it.

The Sitation Lens

The disciplines underneath all of this are clean attribute models, governed syndication, complete coverage and accurate pricing and availability. These are the ones we have been helping brands build for more than a decade. What is changing is the speed and the surface. What is not changing is the underlying work.

Steve Engelbrecht, our CEO and founder, recently described the shift this way: “We, as human workers working with them, can start to think about these AI agents as human‑like colleagues. We can speak to them. We can train them. We can give them feedback.” The same logic applies to how brands prepare their data for those agents. The product information you publish is, in effect, the training material an agent uses to represent your brand in conversations you will never see.

This post is the first of three on agentic commerce. The next two will cover discoverability in an agent‑mediated world and the evolving content contract between brands and retailers. Each one builds on the same idea: in a market where the shopper is increasingly a piece of software, the brands that prepare their product information thoughtfully will be the ones that get found, recommended, and bought.

Ready to Assess Your Data?

If you are not sure how your product information would hold up in front of an AI agent, that is the conversation we like to have. Get in touch to talk through a product data readiness review with our team.