Stripe is treating agentic commerce as the next interface for the internet, and the real fight now is over the plumbing that lets software buy safely.
John Collison's appearance on Bloomberg's Odd Lots arrives at a moment when Stripe is no longer talking about AI shopping as a far-off concept. The company has already moved from theory to products, with Instant Checkout in ChatGPT, the Agentic Commerce Protocol, and a fresh partnership with Google's Gemini, which together show how quickly this market is becoming a race to define the default rails for machine-led transactions. Bloomberg framed the discussion around a simple but disruptive idea: AI agents are moving from helpers that suggest purchases to actors that can discover, negotiate, and complete them on a user's behalf.
That matters because e-commerce was built for human friction. Search, scrolling, comparison shopping, and checkout all assume a person is in the loop at every step. Stripe's case is that this is already changing, and the company's 2025 annual letter said businesses on its network processed $1.9 trillion in volume last year, up 34% from 2024, while Stripe added that agentic commerce is moving into real-world experimentation rather than remaining a slide-deck fantasy.
The hardest part of autonomous commerce is not whether an AI can recommend a product. It is whether the system around it can verify identity, scope permissions, move money, and handle disputes without leaking payment credentials into every layer of the stack. Stripe's new Shared Payment Token is meant to do exactly that, letting an application such as ChatGPT initiate a payment without exposing the buyer's card details, while keeping the token tied to a specific merchant and cart total.
That is only one piece of the problem. Stripe says ACP creates a shared language between merchants and AI agents, so product data, checkout, order submission, tax handling, and fulfillment can work across different interfaces rather than through one-off integrations for every model and app. The company also says merchants can still use other payment providers, which matters because no one wants a fragmented future where every agent requires its own private commerce stack.
Trust is the other missing layer. Payments executives at Stripe have been explicit that the shift is not just about faster checkout, but about rules. Will Gaybrick said agents will spend money faster than people, and Meta's Ginger Baker said payments will move from being a moment to being a policy, meaning users will set guardrails such as limits, permitted categories, and payment preferences instead of approving every single transaction manually. For founders, that means the winning products will not just move money, they will encode intent, permissions, and fraud controls into the transaction itself.
Crypto rails fit into this picture because programmable money and instant settlement are attractive for machine-to-machine systems, especially when micropayments, cross-border flows, and API billing start to matter. Stripe's 2025 letter said it launched machine payments for developers to charge agents directly for API calls, MCP usage, and HTTP requests using stablecoin micropayments, and it also said stablecoin payments volume doubled to around $400 billion in 2025. That does not make stablecoins the only answer, but it does explain why the infrastructure debate is broadening beyond cards and into settlement layers built for software rather than shoppers.
What founders should change
If agentic commerce scales, the biggest strategic shift for startups and D2C brands is that discovery will not be driven only by human clicks. Stripe and OpenAI's Instant Checkout lets users buy inside ChatGPT rather than bouncing out to a merchant site, and Stripe says merchants can now turn AI-driven discovery into a sale inside the conversation itself. That changes the funnel. Product pages still matter, but the first buyer may now be a model reading structured catalog data, not a person landing on a homepage.
That means brands need to think less like media companies optimizing for traffic and more like systems designers optimizing for machine readability. Clean product feeds, precise attributes, structured availability, clear pricing rules, and machine-friendly checkout flows become the new SEO. The merchants that win will be the ones that make themselves easy for agents to parse, rank, and transact with, not the ones that simply buy more ads.
There is also a channel strategy shift hiding inside Stripe's current moves. The company now says agents are becoming a new kind of storefront, and it is rolling out tools for businesses to sell across multiple AI interfaces and protocols with a single integration. That suggests the startup opportunity is not only in building the consumer-facing agent, but in building the middleware that lets brands maintain pricing logic, inventory integrity, identity checks, and post-purchase service across several AI front ends at once.
For D2C operators, the practical takeaway is uncomfortable but clear. If a machine can compare two similar products in seconds, then brand story, fulfillment quality, trust signals, and customer service start to matter more than shallow acquisition tricks. Google's Philipp Schindler said the goal is to remove the grunt work of shopping so consumers can focus on the enjoyable parts, which means merchants will increasingly have to compete on the parts that agents cannot commoditize as easily, including loyalty, service, and reliability.
Stripe is not claiming the transition is complete. In fact, Collison has been careful to frame agentic commerce as something that will unfold in stages rather than all at once. But the company is already building as if the line has been crossed, and the implications are bigger than checkout. The next phase of AI monetization may belong to builders who understand that the real product is not the agent itself, but the trust layer that lets the agent spend on your behalf.
Also read: Mistral CEO says engineers "no longer write a single line of code" and what it means for early-stage hiring • NASA's new AI space chip could help spacecraft think for themselves • ERock tests investor demand for gas-powered AI infrastructure.
