Personal artificial intelligence agents are rapidly evolving from novelty assistants into active participants in the digital economy, increasingly making purchases and financial transactions on behalf of their users. However, given the inherent nondeterministic nature of large language models and autonomous software bots, these systems remain prone to hallucination, misinterpretation, and executing actions outside of user expectations. This technological reality raises significant security concerns regarding the possibility of AI misfiring and executing unauthorized transactions.
To get ahead of this emerging problem, American Express is introducing a new framework designed to capture and authenticate a customer’s purchase intent before an AI agent completes a transaction. The core concept behind the initiative is to ensure that a customer’s explicit purchase instructions are reliably carried throughout the entire transaction lifecycle, from the initial product search to checkout and everything that follows.
Dubbed the "Intent Intelligence Framework," American Express hopes the new system will eventually become an industry standard for secure, AI-powered transactions. The release follows the company’s April launch of Amex’s Agentic Commerce Experiences, commonly known as the ACE Developer Toolkit, which provides a suite of developer services that now incorporates the new intent framework.
Industry experts note that agentic commerce introduces a fundamentally new dynamic to digital retail. "These new transactions will have a consumer, a merchant, and an agent," explained Hannah Bozian, Vice President of Agentic Partnerships and Strategy at American Express, in an interview. She emphasized that the company’s primary goal is to ensure that agentic transactions remain as seamless, trusted, and secure as card members have historically expected from traditional payment methods.
While American Express reported that it has not yet witnessed a measurable increase in transaction disputes specifically attributable to agentic commerce, the financial institution is proactively building out this protective infrastructure as the broader market scales. This proactive stance is heavily supported by consumer sentiment data, which indicates that everyday shoppers are increasingly willing to embrace autonomous shopping tools.
According to a recent Amex Trendex report, 66 percent of surveyed U.S. consumers stated they are likely to use AI agents to research the best price for a product or service over the coming months. Furthermore, 61 percent indicated they are likely to rely on AI for product or service recommendations over the next year. As consumer adoption accelerates, establishing robust guardrails has become a strategic priority for legacy financial institutions looking to prevent fraud and maintain consumer trust.
The mechanics of American Express’s new intent framework begin with a specialized API known as Create Intent. This tool is designed to capture the customer’s precise purchase instructions alongside any granular constraints included in their original request, such as a designated merchant’s name, a maximum budget limit, and specific product preferences. While the exact method of collecting this intent may vary depending on the underlying architecture of the AI model involved, the long-term plan is to make the framework flexible enough to universally support all forms of agentic commerce models.
Once the agent compiles the purchase parameters, the customer must authenticate their identity and formally sign off on the intent captured by the system. Following this verification step, 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, which the AI agent presents at checkout instead of exposing the customer’s actual credit card number. By routing payments this way, American Express ensures that purchases remain secure while strictly adhering to the exact parameters authorized by the user.
Furthermore, the generated Intent ID and associated tokens can be leveraged across various stages of the transaction process, giving companies advanced tools to manage critical operational areas such as fraud prevention and dispute resolution.
"Our approach differs from the implementations of others in that it pulls intent earlier into the transaction flow," Bozian noted. "And this is important because it helps mitigate some risk before the risk occurs."
To illustrate the practical application of this safeguard, Bozian offered a common hypothetical scenario: a consumer instructs an AI agent to purchase a pair of white Nike sneakers in a size 9, but due to a model error or hallucination, the agent submits a purchase request for shoes made by Adidas. Under the Amex framework, the system can instantly intercept and prevent the transaction from proceeding because the agent’s action directly violates the customer’s predefined intent.
Bozian also highlighted how consumer expectations around agent consent are likely to evolve over time. While many consumer-facing agents currently require explicit human confirmation before executing any checkout, American Express envisions a future where agents operate with significantly higher levels of autonomy for routine purchases. For example, a consumer might grant their AI agent ongoing permission to automatically reorder household staples like paper towels or toilet paper—product categories where the primary constraints are limited to price thresholds and package sizes rather than specific brand loyalty or merchant preferences. In these highly autonomous scenarios, having a dedicated technological framework in the loop performing continuous background checks becomes vital to ensuring transactions perfectly match the consumer’s original intent.
Maintaining a transparent, verifiable audit log of what was authorized, by which specific agent, and at what exact time also provides a powerful defense against emerging vectors of fraud conducted through or by autonomous software agents.
American Express is already actively engaged in discussions with various partners across the broader payments ecosystem to implement the ACE initiative. This rollout includes agentic tokens housing Amex payment credentials alongside other foundational components of the developer kit. As the agentic commerce ecosystem continues to mature and expand, the company plans to continuously iterate and develop ACE further to meet the changing demands of digital retail.
Hannah Bozian is scheduled to discuss the company’s new "Intent Intelligence Framework" in greater detail at the upcoming TechCrunch Disrupt Industry Stage, where she will explore its broader implications for the future of fintech, payments, and AI-driven commerce.
