American Express has published a new technical framework designed to help card issuers verify purchases made by artificial intelligence-powered shopping agents, marking another step forward as the financial sector grapples with the rise of autonomous e-commerce.

The payments giant released its Intent Intelligence framework, detailing a standardized method to record a shopper’s purchase instructions before AI agents execute transactions on their behalf. As autonomous shopping tools become more sophisticated—capable of independently browsing, comparing prices, and completing checkouts—financial institutions and merchants are facing entirely new categories of transaction risks, payment disputes, and fraud vectors.

Card issuers can utilize these documented purchase records to more effectively detect fraud, approve legitimate autonomous payments, and resolve disputes when transactions deviate from expectations. For retailers, the system offers a reliable mechanism to determine whether an ordering mistake originated with an autonomous AI agent rather than a standard consumer error or fraudulent merchant activity.

Despite the release of the framework, American Express confirmed that Intent Intelligence does not currently feature any live integrations. Instead, publishing the open framework is intended to encourage other payment networks, financial institutions, and technology providers to adopt its approach, helping to establish a unified industry standard rather than immediately introducing closed proprietary capabilities.

"Providing authorized purchase intent before an agent transacts can give issuers greater confidence that an agent is acting as a customer intended, while also helping issuers proactively manage risk and reduce the chance of costly disputes across the payments ecosystem," said Luke Gebb, head of global innovation at American Express, in the company’s official announcement.

The rollout arrives as major payment infrastructure companies race to develop comprehensive tools for agentic commerce. These systems empower autonomous AI agents to act as personal shoppers, navigating vast digital marketplaces to secure goods and services without requiring constant human intervention at every step of the checkout process.

Yet, despite the convenience promised by autonomous shopping assistants, everyday consumers remain notably cautious about surrendering financial control to algorithms. Recent market research underscores these anxieties. In a survey of 1,000 U.S. adults released by payments company NMI, 58% of respondents expressed worry that an AI shopping agent might mistakenly purchase the wrong item, size, color, or brand. Furthermore, 60% of surveyed consumers voiced concerns that an autonomous agent could easily overspend or exceed their intended budgets.

Although American Express initially introduced Intent Intelligence as part of a broader suite of tools in April, the latest publication makes the technical framework broadly available for external companies to build upon. The stated goal is to help advance a common industry approach for intent management in the rapidly expanding landscape of agentic commerce.

What the new American Express framework means for retailers

For online merchants and traditional retailers alike, AI purchasing mistakes present a complex operational challenge. Under traditional e-commerce models, a merchant might fulfill an order exactly as received via data transmission, only to face a contentious chargeback or dispute because the autonomous agent purchased an item the human shopper never actually requested.

That operational friction is precisely where a permanent record of the shopper’s original instructions, commonly referred to as "purchase intent," becomes invaluable. By meticulously documenting what the customer explicitly approved ahead of time, the framework provides clear evidentiary data regarding whether the AI agent faithfully executed those instructions or malfunctioned during the process.

Intent Intelligence is one of five core services that American Express introduced in April through its Agentic Commerce Experiences (ACE) developer kit. The company emphasizes that this framework differs significantly from current implementation models because it transmits a customer’s authorized purchase intent to the card issuer much earlier in the transaction lifecycle—specifically before an agent executes the final purchase.

American Express noted that Intent Intelligence supports a wide variety of advanced purchase controls. Shoppers retain the ability to establish strict maximum spending limits, set precise expiration dates, and dictate specific merchant preferences. Furthermore, users can specify conditional parameters for when an agent should complete a transaction, such as waiting until an out-of-stock product returns to inventory or executing a purchase exclusively during major shopping events like Cyber Monday.

The framework is also built to accommodate complex, multi-layered purchasing scenarios. For instance, a shopper could authorize an AI agent to purchase optional flight insurance only if the system successfully secures the accompanying airline tickets first. Additionally, it can handle multi-merchant transactions, allowing a single intent record to coordinate purchases across different retailers, such as buying matching shoes and socks from separate online storefronts.

How American Express’s Intent Intelligence framework works

The framework is structurally designed to operate seamlessly across diverse agentic commerce models, regardless of which entity initially captures and processes the shopper’s instructions. Depending on how a particular AI agent architecture handles payment routing, the system can transmit purchase instructions directly to a card issuer. Alternatively, a digital wallet provider or another authorized intermediary can pass the structured instructions along the payment chain.

Why AI shopping errors create new payment risks

Beyond assisting card issuers with purchase verification, the framework directly supports American Express’s broader strategic vision to protect both consumers and merchants from the financial fallout of AI errors.

When the company initially unveiled the ACE developer kit, it outlined plans for a specialized coverage tier known as "Agent Purchase Protection." This proposed coverage would apply to eligible transactions involving registered AI agents that utilize the developer kit and successfully transmit authenticated purchase instructions. For merchants, the primary objective of this protection is to substantially reduce unwarranted chargebacks and administrative disputes.

However, American Express confirmed that this specific protection is not yet active or commercially available, noting that there are currently no live ACE integrations utilizing Intent Intelligence—a mandatory prerequisite for transactions to qualify for the coverage. In the interim, card members retain the standard consumer protections and benefits already tied to their eligible American Express payment cards.

This evolving landscape of responsibility also influences how consumers perceive liability when autonomous transactions go wrong. In NMI’s consumer survey, 34% of respondents stated they would directly blame a retailer if an AI tool completed a purchase before they were fully prepared. Another 24% indicated they would hold their payment provider responsible, while a commanding 62% reported they would place the blame squarely on the AI provider.

Where American Express fits in agentic commerce

American Express is far from alone in developing robust technological infrastructure to secure AI-powered shopping. The broader financial technology sector has moved aggressively to establish standards and APIs for autonomous transactions.

In the spring of 2025, major global competitors Visa and Mastercard launched dedicated offerings designed to connect AI agents directly to their established payment networks. Visa’s Intelligent Commerce program provides developers with comprehensive application programming interfaces and tools to integrate payment capabilities directly into conversational and autonomous AI platforms, working alongside major technology partners such as OpenAI, Microsoft, Anthropic, Stripe, and Samsung. Concurrently, Mastercard introduced Agent Pay, a technology framework enabling AI agents to finalize purchases on users’ behalf across both consumer and business-to-business applications.

Major technology companies have similarly pursued proprietary and open standards for agentic commerce. OpenAI and Stripe co-developed the Agentic Commerce Protocol, allowing AI agents to securely interface with merchants to initiate checkouts. Google subsequently introduced its Agent Payments Protocol, which relies on digitally signed cryptographic records to document shopper authorizations, drawing early adoption from enterprise platforms like Salesforce, Shopify, Cloudflare, and Etsy.

Google expanded its commerce footprint with the Universal Commerce Protocol, an open standard developed alongside Shopify and other industry partners to facilitate direct checkout within AI shopping environments. Major retailers, including Newegg, Wayfair, Walmart, and Macy’s, have since integrated the protocol to allow shoppers to purchase goods seamlessly through AI-driven search and conversational interfaces.