Intro: Amex’s Closed-Loop Bet on Agentic Commerce

American Express is not just participating in the agentic commerce race—it is trying to own the track. With its Agentic Commerce Experiences (ACE) developer kit, Amex has built a closed-loop system that lets AI agents shop and pay on behalf of users, using single-use tokens and intent contracts to enforce transaction boundaries. But the system’s opacity in validation—a deliberate black box—could become a trust liability as enterprises demand auditability.

Luke Gebb, Amex’s EVP and global head of innovation, told VentureBeat: “Some of what is missing so far is the perspective of a company like ours: We feel that trust and security are critical to advancing this space. This is really the first time that an issuer is coming to the table.”

Why this matters: For executives building AI commerce strategies, Amex’s move signals that the payment layer is becoming a competitive moat—but the choice between closed-loop control and open interoperability will define the next decade of autonomous transactions.

The ACE Kit: How It Works

The ACE developer kit provides five integrated services: agent registration, account enablement, intent intelligence, payment credentials, and cart context. The key innovation is the “intent contract”: a user defines what the agent should do, and the system generates an Intent ID and a Proof of Intent Token. When the agent finds the item, it requests a single-use payment token bound to the user’s constraints—for example, a $500 spending limit. “That token won’t allow for a purchase of $600 because it has controls built in,” Gebb explained.

This tokenization gives Amex granular control over each transaction, but the validation process—how the system checks that the agent’s actions match the user’s intent—remains opaque. Amex abstracts how it performs validation, even though it explains at which layer it does it. This black box could hinder trust, especially in high-value or regulated transactions.

Strategic Consequences: Winners and Losers

Winners

  • American Express: By controlling both the issuer and network layers, Amex captures the full transaction value and sets the rules for agentic commerce. It can enforce security standards, reduce fraud, and build a proprietary ecosystem that competitors cannot easily replicate.
  • AI agent developers: ACE provides a ready-made payment infrastructure with built-in trust mechanisms, reducing the need to build custom payment integrations. Developers can focus on agent logic while Amex handles authorization and settlement.
  • Consumers: Single-use tokens bound to intent constraints offer strong protection against rogue agents. If an agent tries to overspend or buy unauthorized items, the token simply fails.

Losers

  • Visa and Mastercard: As open-loop networks, they cannot offer the same level of end-to-end control. If Amex’s closed-loop model gains traction, it could fragment the payment ecosystem and erode their dominance in the agentic commerce segment.
  • Merchants outside Amex’s network: Businesses that do not accept Amex or integrate with ACE risk being excluded from AI-driven transactions that require Amex’s tokenized payments. This could create a two-tier market.
  • Open protocol advocates: Projects like Google’s Agent Pay Protocol (AP2) aim for interoperability, but Amex’s closed-loop approach could slow adoption of open standards if developers prefer the simplicity of a single, secure platform.

Second-Order Effects: What Happens Next

The ACE kit’s black box validation is a double-edged sword. On one hand, it protects proprietary algorithms and reduces attack surface. On the other, it creates uncertainty for enterprises that need to audit agent behavior for compliance or dispute resolution. Raj Ananthanpillai, founder and CEO of identity verification provider Trua, warned: “Without a clear, high-assurance cryptographic link proving that an agent is acting under the explicit authority of a verified human owner, merchants, issuers, and networks face heightened risks of repudiation, massive chargebacks, sanctioned people conducting financial transactions, and fraud.”

If Amex does not open its validation layer to third-party audits, it may face resistance from risk-averse industries like healthcare, finance, and government. Conversely, if it does open up, it could become the de facto standard for agentic commerce trust.

Market / Industry Impact

The shift from human-initiated payments to agent-initiated, intent-bound transactions will redefine payment authorization. Traditional card-present authentication (CVC, 3DS) is inadequate for autonomous agents. Token-based, context-aware authentication—like Amex’s single-use tokens—will become the norm. This could fragment the payment ecosystem into closed-loop (Amex, potentially Visa/Mastercard with proprietary protocols) and open-loop (AP2, Stripe) segments. The winner will be the one that balances security with interoperability.

Executive Action

  • Evaluate your agentic commerce strategy: If your business relies on AI agents for purchasing, consider whether a closed-loop system like ACE offers sufficient flexibility or if open protocols are a better fit for multi-network transactions.
  • Demand transparency: Push payment providers to disclose their validation mechanisms. Opacity may be acceptable for low-value transactions, but for high-value or regulated purchases, auditability is non-negotiable.
  • Monitor regulatory signals: Regulators are watching AI-driven financial transactions. Amex’s closed-loop model could attract scrutiny if it creates anti-competitive barriers. Stay ahead by advocating for open standards that ensure fair access.


Source: VentureBeat

Rate the Intelligence Signal

Intelligence FAQ

It’s a closed-loop system from American Express that lets AI agents shop and pay using single-use tokens bound to user-defined intent contracts, providing end-to-end control over transactions.

Through agent registration, account enablement, intent contracts (Intent ID + Proof of Intent Token), and single-use payment tokens with built-in constraints like spending limits.

Without transparent validation, merchants and regulators cannot fully audit agent behavior, increasing risks of repudiation, chargebacks, and fraud—especially in high-value or regulated transactions.

American Express itself (new revenue stream, stronger moat), AI agent developers (ready-made payment infrastructure), and consumers (enhanced security via tokenized constraints).

As open-loop networks, they cannot offer the same end-to-end control. They may lose market share in agentic commerce unless they develop proprietary protocols or push for open standards that level the playing field.