The Argument
Not all agents are the same. The term "AI agent" has become a catch-all for any autonomous system that acts on behalf of a human, but this conflation obscures the critical differences between agent types - differences in function, in delegation design, in trust architecture requirements, and in the design challenges they present. A comparison agent that evaluates products across merchants operates under fundamentally different constraints than a negotiation agent that haggles on price, or a budget agent that manages spending across categories over time. The essay proposes a taxonomy of commerce agents - a classification system that maps the primary agent types by their function, their delegation design, and their trust architecture requirements.
The Evidence
The essay identifies five primary agent categories. The Comparison Agent evaluates products across merchants against structured criteria - it requires signal clarity and machine-readable product data but operates within narrow delegation scope. The Negotiation Agent engages in price and terms negotiation on behalf of its principal - it requires broader authority, real-time decision-making within constraints, and the ability to make binding commitments. The Budget Agent manages spending across categories and time periods - it requires temporal trust architecture and the ability to balance competing priorities within a financial envelope.
The Subscription Agent manages recurring commerce relationships - renewals, cancellations, upgrades, and downgrades - requiring continuous delegation maintenance and authority drift detection. The Purchasing Agent executes end-to-end transactions from discovery through payment - it requires the broadest delegation scope and the most robust trust architecture. The essay maps cross-cutting design principles that apply across all agent types: every agent requires an operational envelope, every agent must handle failure gracefully, and every agent must maintain legibility to its human principal regardless of its specific function.
The Implication
The taxonomy is not merely an academic classification. It is a design tool. Merchants preparing for agentic commerce must understand which agent types they will encounter and design their infrastructure accordingly. A merchant optimised for comparison agents (structured product data, transparent pricing) may be entirely unprepared for negotiation agents (dynamic pricing APIs, real-time terms adjustment) or subscription agents (recurring delegation verification, authority drift detection). The taxonomy reveals that "becoming agent-ready" is not a single capability but a spectrum of capabilities matched to specific agent types - and that the design requirements for each type are sufficiently different to require distinct infrastructure investments.