AI Agents Are Now Buying on Behalf of Users — Amex's Intent Framework Reveals What Marketplace Trust Must Look Like Next

American Express has launched an 'Intent Intelligence Framework' that captures and authenticates a customer's purchase intent before an AI agent completes a transaction. The system generates a unique Intent ID and a Proof of Intent Token, which replaces the user's actual card number at checkout. This token locks in the buyer's constraints — merchant preference, price ceiling, product specs — before any money moves. Amex is positioning this as a proactive industry standard as agentic commerce scales.

This signals a fundamental shift in who — or what — is the buyer in a transaction. When AI agents execute purchases autonomously, the traditional demand-side of a marketplace is no longer a human making a conscious, real-time decision. That breaks core assumptions marketplaces rely on: informed consent, intentional selection, and traceable accountability — principles central to community marketplace best practices. Amex is effectively defining what 'authorized demand' means in an agentic world. The fact that a major payment network is building infrastructure around this — not reacting to disputes after the fact — tells you the volume of agent-driven transactions is already significant enough to warrant systemic controls.

DEMAND: The buyer is no longer always a person at a screen. AI agents act as demand proxies, and their decision logic may not match actual buyer preference. Marketplaces optimized for human browsing behavior will be mismatched for agent-driven purchasing patterns.

TRUST: Trust in a marketplace has always been bilateral — buyer trusts seller, seller trusts buyer. Agentic commerce adds a third party (the agent) that neither the marketplace nor the seller controls. Without intent verification infrastructure, disputes become unresolvable and trust erodes on both sides.

LIQUIDITY: Agent-driven demand could dramatically increase transaction velocity — agents don't hesitate, comparison shop slowly, or abandon carts. But if agents mismatch supply to user intent, return rates and disputes spike, which destroys liquidity through friction and cost.

ONBOARDING: Marketplaces will increasingly need to think about onboarding agents, not just users. An agent needs to know what your marketplace sells, how pricing works, and what constraints it can apply — before it transacts. Founders revisiting their marketplace launch strategy guide may find that agentic onboarding requirements demand a fundamental rethink of how platform rules are structured and communicated.

MONETIZATION: Agentic buyers will be price-constraint-aware by default. If an agent is instructed to find the lowest price within a category, premium listings or take-rate assumptions built around human anchoring bias no longer hold.

FRAUD: Agent logs create an auditable intent trail — a new form of transaction-level accountability that could reduce fraud disputes. Marketplaces without this infrastructure will absorb higher chargeback costs as agent errors get misclassified as fraud.