Stripe's John Collison on How Agentic Commerce Will Reshape the Internet | Odd Lots

Agentic commerce involves AI systems autonomously purchasing goods or services on behalf of users. This shift manifests in two primary categories: consumer-facing applications where AI researches and executes purchases, and developer-led B2B scenarios where code agents procure resources. The transition from traditional keyword-based search to AI-powered research allows for more nuanced product discovery, enabling users to set specific constraints for furniture, clothing, or technical tools. This evolution benefits smaller merchants by breaking the dominance of traditional aggregators and surfacing niche brands that better align with specific user needs.

Advertising remains a fundamental driver of culture and commerce, even as AI agents take on more decision-making roles. Brand affinity continues to influence human choices, ensuring that advertising retains its value in an agentic world. To facilitate this, businesses must ensure their product catalogs are machine-readable, providing real-time data on stock levels and pricing. Stripe and other infrastructure providers are working to wire up these systems, creating secure, standardized protocols that allow agents to interact with merchants without requiring users to manually re-enter payment details or navigate complex web forms.

A core debate in AI adoption centers on whether to build websites specifically for machine consumption or to rely on AI's growing ability to navigate unstructured, human-centric interfaces. While some companies are developing textual, machine-compatible interfaces, others are betting that AI will become proficient enough to use standard browsers and mouse inputs. Businesses that prioritize agentic-friendly checkout processes are likely to capture early demand, as consumers increasingly prefer the convenience of one-click, agent-assisted transactions over traditional, high-friction web navigation.

The rise of AI agents makes microtransactions economically viable for the first time by removing the human mental load associated with small, frequent payments. Stablecoins offer a potential infrastructure for these tiny, per-request transactions, allowing for seamless access to data or services without requiring individual account relationships. While concerns exist regarding the "end of the free internet," proprietary data remains a significant barrier. Content owners are increasingly licensing their data to AI companies, creating a landscape where high-quality information is protected behind paywalls or licensing deals rather than being freely available for scraping.

Concerns regarding rogue AI agents are mitigated by maintaining a "human-in-the-loop" model, where users retain final approval authority over significant transactions. Corporate resource allocation in the AI era requires a balance between exploiting current capabilities and exploring long-term shifts, with engineering teams showing the fastest adoption rates. Despite the rapid pace of change, there is a measurable increase in economic dynamism, evidenced by a 71% year-over-year growth in new business creation on the Stripe platform. This suggests that lower barriers to entry and increased access to intelligence are empowering more entrepreneurs to launch competing, efficient offerings.