Stripe on Agentic Commerce: Can AI Agents Buy From You?

Podcast/Substack · ~31 min · Listen · (gathered 2026-09-23)

• The core argument is that companies must adapt their fraud, pricing, and onboarding systems for “agentic commerce,” as traditional payment-focused fraud detection is insufficient for AI agents that can cause losses by consuming free credits or trials without making a payment.
• “Fraud rules are a distribution decision”: Decisions about who can use a product, including agents, are often embedded in fraud, pricing, and onboarding processes, not just explicit AI strategy meetings.
• The “Cursor story” highlights this shift, where agents consumed free compute resources, causing losses before Stripe’s payment-focused Radar fraud system could intervene.
• A company’s ability to distinguish good agent users from abusive ones will determine which customers it can afford to serve in an agent-driven economy.
• A spending limit only addresses an agent’s budget; separate judgments are required for an agent’s task quality and permission to share information.