Personal artificial intelligence agents are increasingly stepping into the role of digital assistants that make transactions, purchases, and bookings on their users’ behalf. However, given the inherently nondeterministic nature of large language models and generative AI systems, these autonomous bots are occasionally prone to hallucinations, misinterpretations, and executing actions they were never actually instructed to perform. In the rapidly evolving realm of digital finance, this vulnerability raises a significant possibility of AI misfiring and executing unauthorized transactions that could leave consumers financially exposed.
To get ahead of this emerging problem before it becomes a widespread industry crisis, financial services giant American Express is introducing a new technological framework specifically designed to capture, verify, and authenticate a customer’s true purchase intent before an AI agent is ever allowed to complete a transaction.
The core idea behind the initiative is to ensure that a customer’s explicit purchase instructions are securely carried throughout the entire transaction lifecycle, according to American Express. This protective oversight spans from the initial product search and purchase authorization to everything that follows in the post-purchase and fulfillment stages.
Dubbed the “Intent Intelligence Framework,” the company hopes that this newly minted mechanism will eventually become a widely adopted industry standard for all AI-powered transactions. As commerce shifts from human-driven clicking and tapping to autonomous agent execution, establishing trust and verifiable boundaries will be paramount for widespread consumer adoption.
These new transactions inherently feature a dynamic new triangle involving a consumer, a merchant, and an autonomous agent, explained Hannah Bozian, the Vice President of Agentic Partnerships and Strategy at American Express, during an interview with TechCrunch. She emphasized that the legacy financial institution wanted to ensure that upcoming agentic transactions remain just as seamless, trusted, and secure as the traditional experiences that its card members have grown to expect over decades of service.
This framework’s release directly follows the company’s April launch of Amex’s Agentic Commerce Experiences, commonly referred to as the ACE Developer Toolkit. That developer suite includes a handful of dedicated financial services that now officially incorporate the newly unveiled intent framework as part of their core architecture.
American Express noted that while it has not yet seen a measurable spike in formal consumer transaction disputes specifically driven by agentic commerce errors, leadership wanted to be proactive in building robust security technology well before this type of automated commerce scales to mass markets. Part of this urgency stems from clear market signals showing that everyday consumers are increasingly eager and ready to try out agentic commerce for their everyday shopping and research needs.
In a recent Amex Trendex report published by the company, an overwhelming 66 percent of surveyed U.S. consumers stated that they are likely to use AI agents to research and compare the best prices for products or services over the coming year. Furthermore, 61 percent of respondents indicated that they are likely to utilize autonomous AI agents to secure personalized product or service recommendations over the next 12 months, according to data shared by the company.
To kick off implementation, Amex’s new intent framework begins with a specialized Create Intent API. This application programming interface is engineered to capture the customer’s precise purchase instructions alongside any other specific constraints they may have included in their original request, such as a designated merchant’s name, a strict maximum budget limit, and other granular product preferences like size, color, or material specifications. How this intent is collected from the user may naturally vary depending on the specific underlying AI model or platform involved, but the strategic plan is to make the framework flexible enough to support all existing and emerging agentic commerce models across the digital ecosystem, Bozian noted.
Once the parameters are set, the customer authenticates their identity and formally signs off on the purchase intent captured by their chosen AI agent.
Following this user verification, a unique digital identifier known as an Intent ID is generated to represent the specific intent. Alongside this identifier, a cryptographic Proof of Intent Token is created, which securely bundles the customer’s high-level constraints and approvals. This specialized token is then utilized by systems to obtain a secure digital payment credential that the AI agent will safely present at the digital checkout counter, rather than exposing the customer’s actual, sensitive credit card number or primary account details. In executing the transaction this way, American Express asserts that the purchase will remain fundamentally secure while staying strictly aligned with what the customer originally authorized the agent to do.
This newly established Intent ID and its accompanying cryptographic token can also be leveraged across various distinct phases of the wider transaction lifecycle, helping merchant platforms and financial institutions efficiently manage complex operational areas like proactive fraud prevention and streamlined dispute resolution, according to Amex.
The fundamental design philosophy behind the system sets it apart from other emerging verification methods in the financial technology sector. Our approach differs from the implementations of others in that it pulls the verification of intent much earlier into the transaction flow, Bozian explained. She emphasized that this timing is critically important because it helps mitigate financial and operational risk well before the risk can actually occur in the real world.
To illustrate how the safeguard operates in practice, Bozian offered a hypothetical scenario: a customer instructs their personal AI agent to purchase a pair of white Nike sneakers in a specific size 9. However, due to a model hallucination or misinterpretation, the autonomous agent instead submits a commercial request to purchase shoes from a completely different brand, such as Adidas. In such a scenario, the American Express framework could immediately intervene and prevent the transaction from going through because the agent fundamentally deviated from the customer’s authenticated instructions.
Bozian also pointed out that while many consumer-facing AI agents today routinely ask for explicit human consent before pulling the trigger on a purchase, Amex foresees a future where artificial intelligence agents will inevitably act with a much higher degree of autonomy in routine cases. For instance, a modern consumer might grant their trusted AI agent ongoing permission to automatically re-order household staples and consumables when supplies run low—such as paper towels or toilet paper—product categories where the only real constraints are a maximum price threshold and basic sizing, rather than strict loyalty to a specific merchant or brand. In those kinds of high-frequency, low-friction scenarios, it makes profound technological sense to have an intelligent security layer in the loop performing that extra verification check to ensure the resulting transactions strictly match the customer’s underlying intent.
Maintaining a permanent, auditable log of precisely what was authorized, by which specific agent, and at what exact time could also go a long way toward addressing emerging forms of fraud committed on behalf of or through compromised autonomous agents, she noted.
American Express is already actively engaged in high-level discussions with various commercial partners across the broader payments ecosystem to facilitate the implementation of the ACE suite. This rollout encompasses the agentic tokens containing Amex’s proprietary payment credentials, as well as the various supplementary components included within the developer kit.
The company stated that it plans to aggressively develop the ACE infrastructure further as the wider agentic commerce ecosystem continues to mature and evolve in the coming years.
The broader implications of these developments will take center stage at upcoming industry gatherings. On October 13 at TechCrunch Disrupt 2026’s Industry Stage, Hannah Bozian, the Vice President of Agentic Partnerships and Strategy at American Express, is scheduled to discuss the company’s new “Intent Intelligence Framework” in much greater detail, shedding light on what it ultimately means for the rapidly transforming world of agentic commerce.
