Keep pulling the thread on John Collison.
Stripe is working with Google, Microsoft, Meta, and OpenAI on an "agent-to-commerce suite" to create the programmatic infrastructure for AI-driven commerce.
A core debate at the frontier of AI is whether to make websites and tools more easily consumable by AI, or if AIs will simply become proficient enough at unstructured computer use to navigate the existing human-centric web.
Agentic commerce is expected to make microtransactions viable for the first time because AI agents can handle the mental load of deciding on small, frequent purchases that is too high for humans.
Stripe is building infrastructure to support stablecoins for agentic commerce, which could allow AI agents to pay for services on a per-use basis without requiring users to sign up for each service.
New business creation on Stripe grew 71% year-over-year in Q1, an increase attributed to the "intelligence explosion" from AI enabling more entrepreneurship.
John Collison predicts that AI will lead to greater economic dynamism, resulting in the formation of more, and smaller, firms as the barrier to creating competing products is lowered.
Stripe and Cloudflare have an agreement to allow an AI agent to purchase and register a domain name on behalf of a user.
Stripe categorizes agentic commerce into two main areas: consumer-facing, such as buying through ChatGPT or Claude, and B2B/developer-led, where an agent like Claude Code purchases resources.
AI-powered product research is beneficial for smaller merchants and brands by improving their discoverability beyond traditional aggregators.
A key challenge in enabling agentic commerce is unwinding historical website bot protections and creating a secure system, like a "TSA Pre for bots," to allow approved agents to transact.
John Collison predicts that businesses supporting agentic commerce sooner will outperform those that don't, similar to how third-party sellers on Amazon with Prime are chosen more often.
The dominant economic model in the AI economy is usage-based billing, where users pay per token or per usage, because there are marginal costs for each AI operation.