Inside AMEX’s agentic commerce stack: How intent contracts and single-use tokens enforce AI transactions

2026-05-04 · Source: VentureBeat · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Cloud Computing & IT Infrastructure · Depth: Intermediate, extended

Summary

American Express (Amex) is developing the Agentic Commerce Experiences (ACE) developer kit, a closed-loop system designed to enable AI agents to shop and pay within its own payment network. This system aims to address critical issues in agentic commerce, such as trust, control, accountability, validation, and security, by providing full transaction control at the payment layer. Unlike other card providers, Amex's unique position as both card issuer and payment network allows it to validate agent-led transactions internally. The ACE kit includes features like agent registration, account enablement, intent intelligence with an "Intent ID" and "Proof of Intent Token," payment credentials with single-use tokens and spending limits, and cart context validation. While Amex emphasizes trust and security, the specific validation processes remain a "black box," raising concerns among practitioners about transparency and widespread adoption.

Key takeaway

For CTOs and VPs of Engineering evaluating agentic commerce solutions, Amex's ACE kit highlights the critical need for integrated trust and security at the payment layer. While its closed-loop system offers robust control and validation within the Amex network, you should scrutinize the transparency of any vendor's validation processes to ensure auditability and prevent "black box" issues. Prioritize solutions that offer clear, verifiable mechanisms for agent identity, intent, and transaction integrity to build consumer and merchant confidence.

Key insights

Amex's ACE kit offers a closed-loop system for AI agents to conduct secure, controlled transactions within its payment network.

Principles

• Trust and security are paramount for agentic commerce adoption.
• Full transaction control at the payment layer is a missing piece.
• Closed-loop systems can enhance validation and accountability.

Method

The ACE kit's workflow involves agent registration, account linking, user-defined intent contracts with proof tokens, single-use payment tokens with spending limits, and cart context validation against original intent.

In practice

• Implement single-use tokens for agent-led payments.
• Define clear intent contracts with verifiable proof tokens.
• Integrate spending limits into agent payment credentials.

Topics

• Agentic Commerce
• AI Agents
• Payment Protocols
• Amazon Web Services
• OpenAI Partnership

Best for: CTO, VP of Engineering/Data, Investor, Director of AI/ML, AI Architect, Consultant

Related on AIssential

• The Race Is on to Keep AI Agents From Running Wild With Your Credit Cards — New FIDO Alliance standards will secure AI agent transactions, preventing hijacking and ensuring user intent.
• From Credit Cards To An AI Concierge: How Amex Ventures Backs Startups Building Autonomous Commerce — Amex Ventures backs agentic commerce systems that deliver seamless, personalized, and secure end-to-end customer experiences.
• Is the Internet Really Ready for Agentic Commerce? — Agentic commerce promises efficiency but introduces significant fraud and integrity challenges requiring advanced identity verification.
• Agentic Electronic Design Automation: A Handoff Perspective — Agentic EDA requires explicit "handoff validity" across design stages, workflows, and organizational boundaries for trustworthy operation.
• Enable safe agentic payments with built-in guardrails using Amazon Bedrock AgentCore payments — Amazon Bedrock AgentCore payments enables secure agentic transactions by enforcing infrastructure-level guardrails against financial risks.
• Visa is handling AI-prompted transactions for OpenAI - but can you trust it? — Agentic commerce requires robust security protocols and user-defined guardrails to mitigate new risks from AI-driven transactions.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by VentureBeat.