The checkout button may soon belong to a machine. As artificial intelligence agents move into booking travel, managing subscriptions and executing purchases on behalf of consumers, the payment layer underneath those transactions is becoming a critical battleground. Buy now, pay later (BNPL) providers are not waiting to find out whether AI agents will remember to offer installment options. They are embedding themselves into the infrastructure now, before the defaults get set.
Both Klarna and Affirm announced separate integrations with Stripe designed to extend their pay-later products into agentic commerce flows. The mechanism in both cases is Stripe’s Shared Payment Tokens, a tool introduced in October that allows AI agents to initiate purchases using a customer’s preferred payment method without exposing sensitive credentials. Klarna’s integration means its flexible payment options will be available at U.S. merchants that already offer Klarna through Stripe, without requiring merchants to do any additional integration. Affirm’s expanded partnership is structured similarly, with the goal of eventually bringing its pay-over-time options into what the companies described as AI-driven commerce experiences.
The urgency behind both deals reflects a structural concern: AI agents, left to their defaults, tend to route transactions through stored card credentials. That pathway works, but it bypasses the financing options that BNPL providers have spent years embedding at the point of sale. Klarna Chief Commercial Officer David Sykes framed it directly: “As AI agents begin purchasing on consumers’ behalf, it’s critical that flexible payment options remain available.”
The Tokenized Layer Underneath Agentic Commerce
Stripe’s Shared Payment Tokens are the connective tissue making these integrations possible. They are designed to be programmable, reusable and interoperable, built for environments where a software agent, rather than a human, is navigating checkout. By integrating this token layer, BNPL providers ensure their products surface as available options even when the buyer is automated.
The Affirm integration is designed to let shoppers see the total cost upfront and choose a repayment plan even when an AI assistant is managing the browsing and purchase process, while allowing merchants to accept those payments on the back end through Stripe. The SPT framework also extends reach: any merchant that offers Affirm, regardless of whether they have a direct Stripe integration, will eventually be able to accept these transactions in agentic flows when supported by the AI platform.
For Klarna, the Stripe partnership is one part of a broader infrastructure push. The company has also expanded its partnership with Google to support the Agent Payments Protocol, an open standard for secure AI-driven payments, and launched its own Agentic Product Protocol to make its products readable by AI discovery systems. The pattern across all of these moves is the same: get into the agent’s decision environment early, at the protocol and token level, before commerce flows calcify around simpler payment defaults.
The Infrastructure Race Is On
The BNPL sector’s urgency is grounded in data about where consumer behavior is heading. PYMNTS Intelligence research found that 49% of interested consumers would allow an agentic AI assistant to complete both routine and larger research-driven purchases under the right conditions. That conditional openness willing to delegate purchasing, but only within recognizable financial guardrails is precisely the dynamic BNPL providers are trying to serve. Consumers who already rely on installment plans want those options available regardless of how they shop.
Earlier PYMNTS Intelligence data shows that 25% of bridge millennials used BNPL in December 2025, up 56% from the prior month, even as overall BNPL usage dipped to 14%. That demographic concentration, younger, digitally native, already accustomed to delegating financial tasks overlaps substantially with the population most likely to adopt AI shopping agents. Losing visibility with that cohort at the agentic checkout would represent a significant erosion of the conversion advantages BNPL has built.
McKinsey projects that by 2030, the U.S. B2C retail market alone could see up to $1 trillion in revenue orchestrated through agentic commerce. For BNPL providers, the question is not whether that market materializes, but whether their products are in the payment option set when it does.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.