American Express (Amex) is building a system that allows AI agents to buy and pay on behalf of users, but for now it’s only within its own payment network and still involves a black box that could hinder trust and auditability.
Amex already participates in agent trading protocol projects, especially Google Agent Payment Protocol (AP2)which focuses on interoperability. Amex’s Agentic Commerce Experiences (ACE) Developer Kit, on the other hand, touches on something most protocols currently lack: full transaction control at the payment layer.
But it’s still not completely transparent as to how it handles validation. ACE uses a closed-loop system, which acts as a card issuer and payment network, to validate agent-directed transactions.
Luke Gebb, executive vice president and global chief innovation officer at Amex, told VentureBeat that the company believes this model is the missing piece in agent commerce.
“Part of what’s missing so far is the perspective of a company like ours: we believe that trust and security are critical to moving this space forward,” Gebb said. “This is really the first time an issuer has come to the table.”
Amex is in that interesting space: Unlike other financial institutions or card providers like Chase or Bank of America, Amex can route transactions through its American Express network. Visa and Mastercard are two of the most well-known payment networks, but these companies do not issue cards themselves and must work with a bank.
The continuing black box of agent trading
The ACE Kit is just one approach to addressing some of the biggest issues in agent trading: trust, control, accountability, validation and security.
Consumers generally don’t want dishonest agents running away with their bank accounts and starting buying things. Merchants don’t want to be left with unpaid items. Banks don’t want to deal with an influx of chargebacks and the possibility of fraud.
Projects like the ACE kit aim to build trust and accountability by verifying an agent’s identity and goals. This can build the trust that agent trading desperately needs.
Amex claims it also offers validation, although the process behind this is unclear. You are abstracting how you do validation, although you explain what layer you do it at. More traditional systems feature a combination of deterministic controls and flexible semantic evaluation that helps match intent and outcome for validation. Amex said agents built with ACE can submit users’ shopping carts and compare them to the agent’s original intent. However, they did not reveal how this works.
Professionals building the agent trading ecosystem lament that despite progress in creating a layer of trust, many black boxes remain that could hinder widespread adoption.
Raj Ananthanpillai, founder and CEO of identity and verification systems provider Trua, told VentureBeat that payment protocols and software kits like Stripe’s Agentic Commerce Suite, Google’s Verifiable Proof of Intent Chain, and the ACE Developer Kit "They excel at handling evidence, verifiable authorizations, and the mechanics of fund movement, but leave prior human validation opaque and underdeveloped."
Ananthanpillai continued: "Without a clear, high-security cryptographic link proving that an agent is acting under the explicit authority of a verified human owner, merchants, issuers, and networks face increased risks of repudiation, mass chargebacks, sanctioned individuals engaging in financial transactions, and fraud."
The ACE team
The ACE Developer Kit solves several operational issues with agent trading, Gebb said, and gives developers access to integrated services:
-
Agent Registration
-
Account Enablement
-
Intent Intelligence
-
Payment credentials
-
Cart context
First of all, it involves the registration of the agent, establishing identity and trust with both the consumer and the company’s agents. When a transaction begins, the agent acting on behalf of the customer and the merchant’s agent can verify each other’s identities and trust that they are dealing with the correct entity.
Next comes account enablement, which links the user’s Amex account to their broker and gives the broker permission to act or, in the case of broker trading, buy something.
Intent intelligence creates what Amex calls an intent contract, where the user defines what they want the agent to do. Once the intention is defined, the ACE system generates a Intent ID and a Proof of Intent Token that definitively accredits the authorization in case of litigation.
Amex handles the actual transaction part, where the user pays for the product using a single-use token. ACE sets the payment credentials used for the transaction, subject to intent and restrictions.
“Once the agent has found the item the customer has ordered, such as red shoes, they will make a call to obtain the payment credentials, which is a token that has the limits that the cardholder has provided,” Gebb said. “So, for example, if they said they only wanted to spend $500, that token won’t allow a $600 purchase because it has built-in controls.”
The last piece is cart context and validation, which Gebb says helps banks and brands match a user’s cart that their agent submitted against their intent.
Amex’s approach shows that for agent commerce to really take off, providers need to understand what systems will allow agents to do and who is ultimately responsible if something goes wrong.
