The Hidden Data Needs of AI Shopping Agents
E-commerce is evolving in ways that most retailers aren’t prepared for. The next wave of online shopping won't happen primarily on websites, it will happen through artificial intelligence agents that browse, compare and recommend products on behalf of consumers. Companies that lack the data architecture to serve these agents will become invisible to an important and growing share of their audience.
A recent survey found that 44 percent of shoppers would allow an AI assistant to browse the web on their behalf, rising to 59 percent for shoppers aged 18-34. The report predicts that AI agents will account for $261 billion in online spending by 2030.
Retailers that have spent years perfecting their websites and user experiences are about to find out that AI agents don't care about any of that. What agents need is perfectly structured, machine-readable data about products, pricing and inventory.
In short: if your company's product data isn't accessible to an AI agent, you'll effectively be shut out of this growing channel.
What Shopping Looks Like When Agents Do the Work
When consumers delegate shopping tasks to AI agents, the browsing experience changes. Instead of visiting websites and scrolling through product pages, shoppers ask for a product, "find me waterproof hiking boots for narrow feet costing under $150," and the agent handles the research.
When instructed, agents ruthlessly drill into details like specifications, availability, delivery windows, and price. An agent doesn’t appreciate a brand's aesthetic or marketing copy. It looks for structured information it can quickly parse and evaluate against the shopper’s criteria.
Brand loyalty works differently.
If products are delivered on time and the quality is consistently good, agents learn to favor those retailers. If inventory is inaccurate and product specifications are incomplete, they’ll look elsewhere. It's a transactional relationship that rewards consistent performance.
The Data Gap
Most retailers aren't ready for this shift because their data infrastructure was built for humans. Product descriptions on websites are often written in colorful prose and product specs are inconsistent or incomplete. Pricing depends on complex offers.
Humans can use their judgment to navigate this ambiguity, but AI agents need precision. They need to know what's in stock now, not that a product "usually ships in 2-3 days."
To build systems that serve agents effectively, you need the following elements:
- Machine-readable product catalogs with standardized attributes across categories. This data must be structured for accurate comparison and filtering.
- Real-time inventory visibility that agents can trust. If an agent recommends a product that's out of stock, it will learn to avoid that retailer.
- Dynamic pricing with clear rules that agents can query programmatically. Complex promotions need to be exposed through programming interfaces, not buried in marketing copy.
- Structured fulfillment data, including delivery windows, shipping options, and service levels, that agents can evaluate against user preferences.
Besides exposing this data, maintaining accuracy at scale will also be critical. An agent shopping for a large business might search for hundreds of products in seconds. If your systems can't handle the load or the data is inconsistent, your company will lose sales.
The Role of Google's AP2 in Payment Infrastructure
Data infrastructure can solve the discovery problem but it doesn't address transaction security. When an AI agent initiates a purchase, how does a merchant verify it has the user's permission? How can it distinguish between a legitimate transaction and an error?
Existing payment systems assume a human is clicking the "buy" button. They use behavioral patterns, device fingerprints, and manual verification to detect fraud. None of these work when an AI agent is making autonomous decisions within parameters the user has defined.
This is where protocols like Google’s Agent Payments Protocol (AP2), become relevant. You can think of AP2 as a system of cryptographically-signed digital permission slips, called "mandates," for every transaction. This creates a tamper-proof audit trail that proves the user's intent from start to finish.
AP2 provides the common language and security layer that allows AI-driven commerce to scale safely. Without it, we could end up with a fragmented mess of competing systems that slow progress.
Building the Right Infrastructure Now
For retailers with the right data infrastructure, agent-based shopping can be an opportunity. It creates a more level playing field where performance matters more than marketing budgets. A smaller retailer with accurate inventory and a fast fulfillment process can compete against larger brands with inconsistent data.
The key is to recognize that your data infrastructure is now customer-facing. It’s the interface through which tech-savvy consumers will increasingly discover your company’s products.
Sean Falconer is head of AI at Confluent, a cloud-native and complete data streaming platform.
Related story: Why Agentic Commerce is Becoming a Retail Imperative
Sean Falconer is head of AI at Confluent, where he works on AI strategy and thought leadership. Sean’s been an academic, startup founder, and Googler. He has published works covering a wide range of topics from AI to quantum computing. Sean also hosts the popular engineering podcasts Software Engineering Daily and Software Huddle. You can connect with Sean on LinkedIn.





