# AI Buyers and AI Coworkers Are Opposite Design Postures ## Claim Designing for an AI that *buys from* you and designing for an AI that *works for* you are opposite UX postures — one points outward to machine-readable infrastructure, the other inward to a reviewable, reversible command center — so the single word "agent" demands two different design approaches. <!-- atomicity: 1 idea ✓ --> ## Evidence **Source**: [Agents, Agents Everywhere — Designing for AI Buyers vs AI Coworkers](https://read.readwise.io/read/01kvdb1gcsaxgccgx1c5gp9mfj) (Automattic Design / Woo) > [!quote] > "The word "agent" is overloaded in 2026 commerce. Two unrelated things share it." The *buyer* side: agents read structured data from the backend, not rendered pages, so the work is machine-readability, an "AI-readiness" score, verified-bot traffic handling, and visibility into invisible sessions — outward-facing infrastructure. The *coworker* side: the work is a command center that surfaces synthesized outcomes for review, exposes the reasoning chain (Observation → Impact → Reasoning) instead of asking for blind approval, and makes "Undo" as frictionless as "Approve" — inward-facing trust and reversibility. ## Scope & conditions - Holds for products that must serve both autonomous external agents (transacting) and internal/collaborative agents (operating the product). - Collapses where only one role exists (a pure API has no coworker UX; an internal-tools product has no buyer surface). ## Links - [[3. Resources/Permanent Notes/Real Standards Adoption Means Implementing the Draft and Pushing the Gaps Upstream]] — applies-in: machine-facing buying infrastructure improves when implementers feed specification gaps back into the shared standard - [[AI & the Future of Knowledge Work MOC]] — applies-in — two distinct agent roles reshape product design - [[The Maker Should Not Grade Its Own Homework]] — extends — the coworker posture's "expose reasoning, don't ask for blind approval" is the design instance of independent verification - [[Local Models Are Sovereignty Insurance Against AI Access Cutoffs]] — contrasts — both are "own the durable layer" moves, but one owns infrastructure, the other owns the model fallback - [[3. Resources/Permanent Notes/Readiness Is Predicted by Behavior Over Time Not Static Attributes]] — applies-in: buyer and coworker systems should infer readiness from observed actions rather than a static identity label - [[3. Resources/Permanent Notes/The CMS Decision Turns on How a Site Is Edited Not What It Contains]] — extends: both tool choices begin with the user's actual operating workflow rather than the artifact's surface category. ## Implications - The buyer posture is the merchant-side mirror of courtneyr.dev agent-readiness (web-bot-auth + content signals): give agents structured, verifiable data. - Open question: do the two postures ever converge, or does every agent product eventually need both surfaces built separately?