Agentic Landmark · The Brief
Briefs on the infrastructure, authorization architecture, and commerce strategy of the agentic web: what brands need so agents can find, evaluate, and transact with them.
Google found websites hiding instructions that tell AI agents to recommend them over competitors, with injection detections up 32 percent in three months. The integrity of agent recommendations is not guaranteed, and a verified agent can be compromised after authorization.
Claude Design is the fourth time in thirty years the tool has stopped being the point. Each transition, production speed became a commodity and judgment became the differentiator. Judgment, not velocity, is the unit worth building a team around.
The Google versus OpenAI race decides which interface consumers use. It does not decide whether your brand is in it. That is settled by whether your product data is structured, machine-readable, and connected to whichever protocol the winner runs on.
Treating agentic commerce as a subset of e-commerce is a category error. E-commerce is a channel; agentic commerce is a mediation model that operates across all of them. Brands that miss this optimize product pages while the layer that decides selection goes unbuilt.
Agentic commerce runs on compute, and compute runs on power the grid was not built to deliver. By 2030 AI data centers may out-consume heavy industry, and in Virginia households, not hyperscalers, are already paying for it. That is worth caring about.
Adobe's trillion-visit data settles it: AI-referred traffic now converts 42 percent better than every other channel, an 80-point swing in a year. Yet retail product pages average 66 percent machine readability, so a third of your data is invisible to it.
Fewer than 10 percent of enterprises have scaled AI agents, and the blocker is data, not models. The same data gaps that stall internal agents make a brand invisible to the external agents shopping for consumers. The field is wide open.
The shopping cart is a thirty-year-old skeuomorph nobody questioned, and in agent-mediated commerce it disappears: no gathering, no abandonment, no recovery program to run. Optimizing cart completion is optimizing the last step of a process being bypassed at the first.
McKinsey's twelve AI themes are the right principles, but for commerce the urgent work concentrates in two: structured product data that decides whether agents can select you, and the authorization surface where consumer trust in an agent is won or lost.
BCG mapped four futures for agent-driven retail, then named the imperatives that hold across all of them: discoverability and desirability both reduce to structured, machine-readable product data, so the infrastructure bet is the same in every scenario.
