Agentic Commerce, per MIT Tech Review: Why AI Agents Need Data, Not Banners

AI agents are moving from returning lists of links to actually carrying out tasks: booking, buying, comparing offers. But for this to work reliably, they need two key resources: trustworthy structured data and full user context.

From suggestion to action

Imagine telling a digital agent: "book a family trip to Italy, use my bonus points, stay within budget and pick hotels like the ones we liked before." Instead of returning a list of links, the agent assembles the itinerary and completes the purchase itself. This shift, from assisting to executing, is the essence of agentic commerce.

MIT Technology Review notes that we are on the threshold of the next wave of e-commerce, in which the main interface is neither a website nor an app, but an agent. The change looks evolutionary, but the consequences for business are radical.

The shift is now being accelerated by OpenAI Operator, Google Shopping and the agentic modes of Claude. The technology is already ready; the only question is whether retailers and services are ready for it.

Truth matters more than looks

Traditional e-commerce is optimized for humans: attractive product cards, emotional banners, visual personalization. None of this is relevant to an agent. An agent needs structured data: accurate prices, real stock levels, machine-readable delivery and return terms. It has no use for a "brand story" on the home page; it needs a reliable API or a clean data feed. If the data is outdated or contains errors, the agent will either make the wrong decision or refuse to work with that source altogether.

• Real-time accuracy of prices and availability is a mandatory requirement
• Structured product attributes matter more than marketing descriptions
• Machine-readable delivery and return terms must be directly accessible
• Inaccurate data leads to erroneous transactions or an outright refusal by the agent

This changes investment priorities: money goes not into UX design but into data quality and availability. Companies that keep investing only in visuals risk becoming invisible to agents.

Context drives the choice

The second component is user context. An agent must know not only what to buy, but for whom, under what constraints and with which preferences in mind. An effective agentic experience is built on accumulated history: which hotels the user liked before, which budget is comfortable, which airlines to avoid. This requires the user to trust the system, and a serious approach to managing personal data. An agent that lacks context will fall back to the same behavior as an ordinary search engine.

"Agentic commerce runs on truth and context": this is how MIT

Technology Review formulates the fundamental condition of the new era of shopping.

For companies this means a new task: it is not enough to build a good product; they must also actively structure their offering so that agents interpret it correctly and can explain to the user why this particular option fits better than others.

New rules for business

The move to agentic commerce breaks the familiar sales funnel. Previously, a brand spoke directly to the consumer through advertising, content and UX. Now an agent stands between them, and it decides which supplier makes the user's final shortlist.

• SEO gives way to agent-first optimization: structured data and open APIs
• Brands compete not for the click but for the agent's trust and the quality of their data feeds
• Marketing budgets are being redistributed toward data infrastructure
• Companies that are first to adapt their data for agentic consumption will gain a disproportionate advantage

What this means

Agentic commerce is not a distant future: the tools already exist, major platforms are actively rolling them out, and the first cases with real transactions are already being published. Brands that learn to speak the language of agents (structured, precise, real-time) will win the next era of e-commerce. Those that keep optimizing only for the human eye risk being left behind.