- American Express unveiled the Intent Intelligence Framework on October 13, 2026, to verify what a customer wants before an AI agent pays.
- Amex’s Trendex report says 66% of U.S. consumers are likely to use AI agents to research the best price.
- The agent receives a tokenized payment credential, so it never sees the real card number.
- Amex hopes the approach becomes an industry standard and is talking with unnamed payments-ecosystem partners.
Two out of three. That is roughly the share of U.S. consumers (66%) who told Amex’s Trendex survey they are likely to let an AI agent research the best price for a product or service. Research is the easy part. The hard part starts when the agent is allowed to pay, because a wrong purchase is a real charge, a real dispute and a real loss of trust. On October 13, 2026, American Express introduced a framework aimed at exactly that gap.
What Amex Announced
The Intent Intelligence Framework
According to TechCrunch, the Intent Intelligence Framework is designed to capture and verify a customer’s purchase intent before an AI agent completes a transaction. It extends the Agentic Commerce Experiences (ACE) Developer Toolkit that Amex launched in April. Hannah Bozian, Amex’s VP of Agentic Partnerships and Strategy, said of the new transactions: “These new transactions will have a consumer, a merchant, and an agent.”
Her point is that payments used to be a two-party relationship. Adding an agent creates a third actor that can misread instructions, so the system needs a record of what the human actually authorized.
Trend Insight — The competitive question in agentic commerce is shifting from who builds the smartest shopping agent to who builds the most trusted checkout layer. Whoever owns the proof of intent owns the dispute process.
How the Flow Works
From instruction to payment
The process starts with a Create Intent API that captures the customer’s instructions and constraints, such as merchant name, maximum budget and product preferences. Amex says the way this is collected may vary by AI model. The customer then authenticates and approves the captured intent.
Next, the system issues a unique Intent ID and a Proof of Intent Token that carries the customer’s high-level constraints. That token is used to obtain a digital payment credential, so the agent can pay without ever seeing the customer’s actual card number. The Intent ID and token can later support fraud prevention and dispute resolution.
The sneaker test
Amex gives a simple example. If a customer asks for white Nike sneakers in size 9 and the agent tries to buy Adidas shoes, the transaction could be blocked. Bozian added: “Our approach differs from the implementations of others in that it pulls intent earlier into the transaction flow.”
Trend Insight — Checking intent before payment, rather than disputing after it, moves the control point upstream. For developers, that means agents will need a way to hand a structured, user-approved mandate to the payment network.
Why It Matters for Builders
More autonomy, with an audit trail
Amex expects agents to act more independently in low-risk categories, such as reordering household staples where only price and size matter. It also points to an audit log of what each agent was authorized to do, and when, as a tool against agent-related fraud. The company says it has not yet seen an increase in agentic commerce disputes, which suggests this is preparation rather than a reaction to a crisis.
For startups building shopping or procurement agents, the practical takeaway is to design for explicit constraints from day one: budget caps, merchant allowlists and product specifications that can be serialized into a verifiable token. Amex says it is in discussions with unnamed partners across the payments ecosystem about implementing ACE, so details on adoption remain open.
Trend Insight — Expect every major card network to publish its own intent standard. Teams that abstract the mandate layer now will switch cheaply later.
Related
- Meta Finally Built a Frontier Model. Read the Fine Print.
- Anthropic’s New Model Costs 90% Less. What’s the Catch?
- Satya Nadella Just Said the Quiet Part About AI
- Enterprise AI agent trust drop (75% to 56%)
Sources
- TechCrunch — Amex introduces a framework to prevent AI agents from making unauthorized purchases
- TechCrunch — Artificial Intelligence category
AI Biz Insider · AI Trends EN · aibizinsider.com
답글 남기기