American Express (Amex) has launched a new framework to address risks that artificial intelligence (AI) agents may pose during transactions. As AI agents increasingly complete transactions on behalf of users, AI bots — due to their inherent uncertainty — are prone to acting beyond expectations, which in payment scenarios could lead to erroneous operations or unauthorized transactions.
Amex Launches Intent Intelligence Framework to Enhance Transaction Security
According to Amex, this system, called the "Intent Intelligence Framework," aims to ensure that customers' purchase instructions are fully transmitted throughout the entire transaction lifecycle, from order placement through every subsequent step. The company hopes this framework will become an industry standard for AI-driven transactions.
Amex Vice President of Agent Partnerships and Strategy Hannah Bozian told foreign media in an interview that this new type of transaction involves three parties simultaneously: consumers, merchants, and agents. She noted that the company wants to ensure agent-based transactions remain "seamless, trusted, and secure, meeting cardmembers' expectations." The launch of this framework follows Amex's release of the Agentic Commerce Experience (ACE Developer Toolkit) in April of this year; the toolkit includes multiple services that have now been incorporated into this intent framework.
Amex stated that it has not yet seen an increase in dispute cases due to agentic commerce, but given the growing scale of such transactions, the company wants to establish technical infrastructure in advance. According to data provided by Amex to media, in the recent Amex Trendex report, 66% of surveyed U.S. consumers said they would likely use AI agents to research the best prices for products or services; 61% of respondents said they would likely use AI agents to obtain product or service recommendations within the next 12 months.
The intent framework process begins with the Create Intent API, which captures the customer's purchase instructions and any associated constraints, such as merchant name, maximum budget, and other product preferences. Bozian said the method of intent capture may vary depending on the AI model involved, but the plan is to make the framework flexible enough to support all agentic commerce models. Next, the customer must undergo identity verification and confirm the intent captured by the agent.
The system then generates an Intent ID as a unique identifier for that intent, along with an intent proof token containing the high-level constraints set by the customer. This token will be used to obtain a secure digital payment credential, which the agent uses at checkout instead of the customer's actual credit card number. Amex noted that this keeps transactions secure while maintaining consistency with the actions the customer authorized the agent to perform.
Bozian said: "Our approach differs from other companies' implementations in that we bring intent into the transaction flow earlier. This is critical because it allows us to mitigate a portion of risk before it actually materializes." She gave an example: if a customer asks an agent to buy a pair of white Nike sneakers in size 9, but the agent submits a request to purchase Adidas shoes, Amex can block the transaction because the agent made an error.