J.P. Morgan's "Trillion-Dollar Question"—What Payment Infrastructure Looks Like When AI Agents Shop for You

The Era of AI Agents Shopping on Your Behalf

AI Agents for Shopping - Trillion Dollar Question | J.P. Morgan

Learn about the future of commerce—AI agents that can shop on your behalf—in this interview about Visa Intelligent Commerce.

J.P. Morgan's payments publication "Payments Unbound" has released a comprehensive report analyzing the future of AI agent shopping. Starting from the question "Would you let an AI agent shop for you?", the report details the movement toward agentic commerce—a form of commerce where AI agents search, compare, and purchase products on behalf of humans—spanning startups, tech giants, and payment companies.

At the core of the article is an interview with Visa's Global Head of Growth, Rubail Birwadker. It covers the technical mechanics of "Visa Intelligent Commerce (VIC)" and the fundamental challenge of confirming purchase intent when AI agents make payments.

Industry Background and Trends

Agentic commerce has been rapidly materializing from late 2025 through 2026. McKinsey's October 2025 analysis estimated that $3-5 trillion in global retail spending could be redirected to agentic commerce by 2030. Morgan Stanley also predicts that 10-20% of US e-commerce sales ($190-385 billion) will flow through agents by 2030.

Consumer behavior is also shifting noticeably. According to Morgan Stanley's research, approximately 23% of Americans have made some form of purchase through AI in the past month. Adobe's data shows that traffic from generative AI browsers and chat services to US retail sites increased 4,700% year-over-year as of July 2025.

Google and Amazon are also making moves. Google has implemented agentic shopping and checkout features, while Amazon has expanded Rufus with a "Buy for Me" function. Shopify has begun offering agentic infrastructure capable of cross-merchant cart building.

Visa Intelligent Commerce and the "Intent Confirmation" Challenge

The most notable aspect of J.P. Morgan's article is the detailed explanation of Visa Intelligent Commerce (VIC). VIC is an API set that enables AI agents to access Visa's payment services using tokenized credentials. According to Birwadker, partner companies are already working on integrations, with Visa collaborating with over 100 partners worldwide. More than 30 are developing in the VIC sandbox, over 20 agents are pursuing direct integrations, and hundreds of real transactions have been completed.

However, a fundamental challenge in payments has also emerged. In traditional e-commerce, a "click" clearly signals purchase intent. But in natural language interactions, the boundary between "I like this" and "buy this for me" is ambiguous. Birwadker indicated that in the early stages, agents will seek consumer confirmation before purchases, gradually relaxing controls as usage patterns accumulate.

The question of "liability" also remains unresolved. If an agent makes an unintended high-value purchase, who bears the loss—the user, the merchant, or Visa? Birwadker candidly acknowledged that "much consideration is still needed."

Industry-wide, Visa and over 10 partners announced the "Trusted Agent Protocol" in October 2025. This is an open framework built on existing web infrastructure that enables merchants to distinguish between malicious bots and legitimate AI agents. Mastercard has also announced "Agent Pay", advancing integration with Microsoft's Copilot Checkout. With OpenAI, Cloudflare, and PayPal also involved, payment giants are moving in unison toward agent readiness.

FIS has also announced issuing bank solutions for agentic commerce, planning to launch identity verification using "Know Your Agent (KYA)" data by Q1 2026.

Impact and Implications for EC Operators

The maturation of agentic commerce demands three practical responses from EC operators.

Payment infrastructure readiness: Adapting to Visa Intelligent Commerce and Mastercard Agent Pay is unavoidable in future EC platform design. Specifically, support for tokenized credentials and the ability to process agent-initiated API purchase requests are essential.

Product data structuring: AI agents read structured data, not product page HTML. The ability to provide product names, prices, specifications, and inventory information via APIs or feeds directly translates to capturing agent-driven sales.

Redefining trust and brand building: In a world where agents handle purchases, the effectiveness of traditional SEO and listing ads changes. Morgan Stanley analysts note that "the entire EC funnel is being disrupted," necessitating new brand strategies designed to be selected by AI agent recommendation algorithms.

Visa predicts that millions of consumers will make purchases through AI agents by the 2026 holiday season. Delays in preparation directly translate to missed opportunities.

Summary

What J.P. Morgan's report reveals is that agentic commerce is no longer at the conceptual stage—implementation is progressing at the payment infrastructure level. Payment infrastructure companies like Visa, Mastercard, and FIS are simultaneously advancing agent readiness, while research firms like McKinsey and Morgan Stanley are projecting trillion-dollar market opportunities.

The key watch point going forward is when industry standards for "purchase intent confirmation" and "liability allocation" will be established. Once these are in place, agentic commerce adoption will accelerate dramatically. For EC operators, 2026 will be the inflection point where the gap between early movers and laggards becomes clearly visible.