As personal AI agents increasingly step into the role of digital shopping assistants, making transactions and purchases on behalf of their users, a persistent technical challenge looms large. Because of the inherent nondeterministic nature of large language models, AI bots are prone to hallucinating, misinterpreting prompts, or simply executing actions they were never meant to perform. In the rapidly evolving landscape of digital commerce, this unpredictability raises a significant and costly possibility: AI agents misfiring and executing unauthorized financial transactions.
To get ahead of this growing problem before it becomes an industry-wide crisis, American Express is introducing a robust new framework designed to capture, verify, and authenticate a customer’s exact purchase intent long before an AI agent actually completes a transaction.
The core philosophy behind the initiative is simple yet critical, according to representatives from Amex: there must be a dependable way to ensure that a customer’s initial purchase instructions are faithfully carried throughout the entire lifecycle of a transaction, stretching from the initial product search all the way through to payment and post-purchase management.
Dubbed the “Intent Intelligence Framework,” the financial services giant hopes the new system will eventually become a universally adopted industry standard for the burgeoning realm of AI-powered transactions.
“These new transactions will have a consumer, a merchant, and an agent,” explained Hannah Bozian, Amex’s Vice President of Agentic Partnerships and Strategy, in an interview. She emphasized that the company wanted to ensure that agentic commerce remains just as seamless, trusted, and secure as traditional card members have always expected it to be.
The release of this new framework follows closely on the heels of the company’s April rollout of Amex’s Agentic Commerce Experiences, commonly referred to as the ACE Developer Toolkit. The toolkit initially included a handful of developer services and has now been expanded to incorporate the new intent framework.
Amex noted that it has not yet witnessed an abnormal spike in customer disputes directly caused by agentic commerce. However, the company deliberately chose to adopt a proactive stance, building out this critical infrastructure now as this unique type of automated commerce begins to scale. That forward-looking approach is heavily underscored by clear market signals showing that everyday consumers are increasingly eager to experiment with agentic commerce.
According to findings from a recent Amex Trendex report, 66% of surveyed U.S. consumers stated they are likely to utilize AI agents to research the best possible prices for products or services over the coming months. Furthermore, 61% of respondents indicated they are likely to rely on AI agents for personalized product or service recommendations over the next year.
To kick off its deployment, Amex’s new intent framework begins with a specialized Create Intent API. This application programming interface is engineered to capture a customer’s explicit purchase instructions alongside any secondary constraints they may have included in their original request—such as a specific merchant’s name, a strict maximum budget, and other precise product preferences. While the exact method of how this intent is collected may vary depending on the underlying AI model involved, Amex’s overarching strategy is to make the framework flexible enough to seamlessly support all existing and emerging agentic commerce models, according to Bozian.
Once the instructions are gathered, the customer authenticates their identity and officially signs off on the specific intent captured by the AI agent.
Following this user sign-off, a unique identifier known as an Intent ID is generated, accompanied by a Proof of Intent Token that encapsulates the customer’s high-level constraints. This token is subsequently utilized to obtain a secure digital payment credential. Instead of handing over the customer’s actual, sensitive credit card number to the AI agent or the merchant, the agent uses this temporary digital credential at checkout. By utilizing this method, Amex ensures that the purchase remains thoroughly secure while strictly aligning with what the customer originally authorized the agent to do.
Beyond checkout, this Intent ID and its associated token can be leveraged across various stages of the transaction process, giving companies valuable tools to better manage areas such as advanced fraud prevention and streamlined dispute resolution.
“Our approach differs from the implementations of others in that it pulls intent earlier into the transaction flow. And this is important because it helps mitigate some risk before the risk occurs,” Bozian explained.
To illustrate how the system functions in practice, Bozian offered a common hypothetical scenario: a customer instructs their personal AI agent to purchase a pair of white Nike sneakers in a size 9. However, due to a model hallucination or misinterpretation, the agent mistakenly submits a request to purchase a completely different pair of shoes from Adidas. In a situation like this, Amex’s framework would be able to intercept the error and prevent the transaction from going through because the agent fundamentally violated the user’s established intent.
Bozian also pointed out that while the vast majority of AI agents today still prompt users for explicit consent right before executing a purchase, the financial industry must prepare for a future where agents act with a much higher degree of autonomy in routine scenarios. For example, a consumer might grant their AI agent standing permission to automatically reorder household staples—such as paper towels or toilet paper—categories where the only real constraints are basic price thresholds and specific product sizes rather than brand loyalty or a particular merchant. In these highly automated scenarios, having specialized technology permanently in the loop to perform continuous background checks is essential to ensuring that transactions faithfully match the customer’s intent.
Maintaining a comprehensive digital log detailing precisely what was authorized, by which specific agent, and at what exact time could also go a long way toward addressing potential fraud originating from or involving the agent itself, she noted.
Amex is already actively engaged in discussions with partners across the broader payments ecosystem regarding the implementation of the ACE toolkit. This includes integrating agentic tokens that securely house Amex’s payment credentials, as well as adopting other core components of the developer kit.
The company stated that it plans to continue aggressively developing the ACE platform as the broader agentic commerce ecosystem matures and changes over the coming years.
Industry professionals looking to dive deeper into these developments will have an opportunity to hear more directly from leadership. On October 13 at TechCrunch Disrupt 2026’s Industry Stage, Hannah Bozian, Vice President of Agentic Partnerships and Strategy at American Express, is scheduled to discuss the company’s new Intent Intelligence Framework in greater detail, shedding light on what it ultimately means for the rapidly expanding world of agentic commerce.
