How did Stripe's use of AI evolve, and when did foundation models become critical?

Stripe has used machine learning for over a decade, primarily for fraud detection and internal operations. The shift to AI, specifically LLMs, became a focus when GPT-3.5 emerged (late 2022/early 2023). In the last 1.5 years, Stripe realized the need for its own domain-specific foundation models to create richer, denser payment embeddings for various applications.

How does Stripe's foundation model enhance fraud detection beyond traditional methods?

Traditional ML models struggle with sophisticated fraud like card testing hidden in large transaction volumes. Stripe's foundation model transforms each charge into a dense embedding, allowing the detection of subtle clusters of anomalous behavior in real-time (under 100 milliseconds), uncovering new fraud vectors and suspicious transactions that traditional methods miss.

What is 'friendly fraud' and why is it a significant challenge for AI companies?

Friendly fraud includes non-payment abuse like free trial abuse or refund abuse, where legitimate credentials are used but no genuine revenue is generated. For AI companies, this is existentially threatening because GPUs and inference costs are high, making the marginal cost of services substantial, unlike traditional SaaS where marginal costs were near zero.

How is Stripe adapting its offerings to support AI-native business models?

Stripe provides payments, billing suites, optimized checkout, and fraud protection adapted for AI companies. They've introduced 'token billing' to enable real-time pricing based on fluctuating LLM inference costs and support outcome-based billing, stablecoins, and Link for global reach and efficient transactions, especially given AI companies' rapid global expansion.

What is the Agentic Commerce Protocol (ACP) and what is its significance?

The ACP is a shared standard developed with OpenAI that allows businesses to interact with AI agents, enabling agents to make purchases on behalf of users. It standardizes how product catalogs, pricing, and brand appear to agents, and includes a shared payment token for secure transactions, addressing the challenge of 'good bots vs. bad bots.'

How does Stripe approach the 'build vs. buy' decision for its internal tools, especially with AI?

Stripe's strategy is 'buy and build,' prioritizing buying existing solutions if they meet needs, then considering building if no suitable option exists. They encourage re-evaluating existing builds against new market offerings and favor solutions that integrate with multiple models and providers, like open-source options, to avoid vendor lock-in.

How do LLMs influence deep thinking and the 'social contract' of work at Stripe?

LLMs can produce enticing output quickly, but the speaker argues that deep thinking is fostered by the arduous process of writing and structuring arguments. While LLMs are great for generative tasks and rapid experimentation, over-reliance can lead to 'LM slop' and hinder the development of intellectual 'depth' if not coupled with critical human oversight and validation.

What is Stripe's internal tool 'Hubert' and what challenges does it face?

Hubert is an internal AI tool that converts natural language queries into SQL to ask questions about business data stored in Hubble. While effective for well-structured data, its main challenge is data discovery (identifying the correct tables and fields) in complex, less-documented datasets. Stripe is working on deprecating low-quality tables and improving documentation.

Are AI companies currently in a 'bubble,' according to Stripe's data?

Stripe's data shows AI companies are growing 2-3 times faster than previous SaaS cohorts, are significantly more global from day one, and have strong unit economics. While per-company customer retention might be slightly lower due to a competitive market, customers still value the problem-solving aspect. The speaker suggests that while expectations can be high, the underlying businesses show real value and significant growth potential, especially as inference costs decrease.

How does AI impact GDP and economic growth?

AI should make markets and commerce more efficient, expanding consumption and accelerating business creation. While the exact impact on GDP is debatable, increased efficiency in both production and consumption should lead to meaningful GDP growth. The speaker highlights 'tiny teams' (high revenue per employee) as a key indicator of AI-driven efficiency.

Key Moments

The Agents Economy Backbone - with Emily Glassberg Sands, Head of Data & AI at Stripe

Latent Space Podcast

Science & Technology5 min read98 min video

Oct 30, 2025|4,242 views|91|5

Save to Pod

Key Moments

TL;DR

Stripe leverages AI for financial infrastructure, focusing on fraud, payments efficiency, and enabling agentic commerce.

Key Insights

Stripe processes a significant portion of global GDP, offering a unique vantage point for AI and ML applications in finance.

AI and LLMs are crucial for Stripe's evolution, moving beyond traditional ML for fraud detection to powering complex financial products and agentic commerce.

Agentic commerce, exemplified by the Agentic Commerce Protocol (ACP) with OpenAI, aims to standardize how businesses interact with AI agents for buying and selling.

Stripe is actively developing solutions for emerging AI business models, including usage-based and outcome-based billing, and token billing to track inference costs.

The company prioritizes building robust financial infrastructure for AI companies, enabling global reach and efficient monetization through tools like Link and their billing suite.

Stripe is investing in AI for internal operations, citing significant productivity gains in areas like payment method integration and data analysis.

STRIPE'S MISSION AND AI'S ROLE

Stripe's core mission is to build financial infrastructure for the internet, evolving from payments to encompass a broad range of business needs like subscriptions, tax, and money movement. Processing approximately 1.3% of global GDP ($1.4 trillion annually), Stripe possesses a vast dataset that fuels its AI and Machine Learning initiatives. These initiatives aim to enhance payment experiences, reduce fraud, optimize authorizations, and improve customer-facing features. The company's AI and Data organization spans foundational layers like data platforms and ML infrastructure to applied AI solutions, including a dedicated experimental projects team for novel initiatives like agentic commerce.

EVOLUTION OF AI AND ML AT STRIPE

Stripe has utilized Machine Learning for over a decade, initially for fraud Detection (Radar) and internal operations. The advent of advanced LLMs like GPT-3.5 spurred a significant investment in AI infrastructure and applications, leading to the development of domain-specific foundation models. These models allow Stripe to move beyond single-task ML models to richer, denser payment embeddings that power various downstream applications. This evolution ensures Stripe can handle massive transaction volumes (50,000 transactions per minute) with low latency, enabling sophisticated fraud detection, such as identifying subtle card testing attacks by analyzing transaction sequences as 'language data'.

TACKLING NEW FORMS OF FRAUD IN THE AI ECONOMY

The rise of AI has introduced new economic challenges, particularly 'friendly fraud,' which includes free trial abuse and refund abuse. Unlike traditional SaaS, AI businesses face existential threats from these practices due to the high marginal cost of compute and inference. Stripe is developing specialized radar extensions to combat this, moving beyond detecting stolen credentials to addressing non-payment abuse and economic inefficiencies. The complexity is amplified by high-value enterprise plans and instances like 'free trial cards' designed to bypass payment processes, necessitating advanced solutions that can identify and mitigate these evolving fraud vectors.

ENABLING AI BUSINESS MODELS WITH FINANCIAL INFRASTRUCTURE

Stripe serves as the 'skeletal system' for AI companies, with all Forbes AI50 companies monetizing through Stripe. These lean, global-first businesses leverage Stripe for payments, global reach (median AI company in 55 countries in year one), fraud protection, and flexible monetization. As AI companies often act as 'wrappers' for underlying LLMs, they face pricing volatility. Stripe's 'token billing' addresses this by allowing real-time tracking and pricing of inference costs, providing adaptability to fluctuating LLM expenses. Furthermore, Stripe's billing suite supports various models, including usage-based and outcome-based billing, catering to the dynamic pricing needs of AI services.

THE AGENTIC COMMERCE PROTOCOL (ACP)

A significant strategic move is Stripe's Agentic Commerce Protocol (ACP), developed in collaboration with OpenAI. ACP establishes a shared standard for businesses to interact with AI agents, transforming how financial infrastructure operates for automated purchasing. It standardizes exposing product catalogs, inventory, and pricing to agents and introduces a shared payment token for secure credential transfer, mitigating risks for agents. ACP also aims to differentiate 'good bots' from 'bad bots,' essential for managing demand in dynamic markets. This protocol is designed to be inclusive, working with any payment provider and supporting various agents beyond just OpenAI's ecosystem.

INNOVATIONS IN PAYMENT AND MONETIZATION FOR AI

Beyond ACP, Stripe is facilitating new payment methods and monetization strategies for AI companies. Stable coin adoption is growing, particularly for global reach and high-value transactions where traditional international card fees can be prohibitive. Stripe also sees traction with Link, its consumer product, enabling one-click checkout for dense networks like AI platforms. Internally, Stripe is enhancing its own infrastructure, using AI for tasks like integrating new local payment methods efficiently. The company is also exploring AI-powered monetization tools, such as 'claimable sandboxes' within developer platforms like Vercel, to streamline business creation and payment integration.

THE ECONOMICS OF THE AI BOOM AND FUTURE OUTLOOK

Stripe's unique position allows observation of economic trends. AI companies on Stripe are growing revenue two to three times faster than previous SaaS cohorts and are significantly more global. While per-company retention might be slightly lower, customer stickiness is high as users efficiently switch between competitive AI solutions. The discourse on AI's economic impact suggests a massive potential for expanded commerce, especially by removing time constraints for high-income individuals. Stripe views its role in driving efficiency and enabling new business creation, ultimately contributing to global GDP growth, rather than a speculative bubble.

BUILDING THE FUTURE: INTERNAL TOOLS AND EXTERNAL STRATEGY

Stripe navigates a 'build vs. buy' strategy, often starting with building internal solutions (like early LLM access) and evolving to leverage external products as the market matures. This approach includes programs like 'spotlight' for evaluating and acquiring external solutions and collaborative open-source efforts. The company emphasizes the importance of design and brand in the AI era, noting that a strong user experience and compelling brand narrative are critical differentiators beyond pure technology. This focus on quality extends to obsessing over micro-details in user experience and maintaining rigorous standards, even at scale.

Mentioned in This Episode

●Products

●Software & Apps

●Companies

●Concepts

●People Referenced

Common Questions

Stripe aims to build financial infrastructure for the internet, and AI is increasingly integral to this mission. It involves using data to understand the economy, feed insights back into products for better payment experiences, and building economic infrastructure specifically for AI businesses.

Topics

Stripe Financial Infrastructure Fraud Detection Agentic Commerce Token Billing Global Markets

Mentioned in this video

companySam's Club

Signed up to make its inventory purchasable through ChatGPT and the Agentic Commerce Protocol.

personArvin Shinivas

Mentioned as someone who started as an 'LLM wrapper' and found product-market fit without needing to build underlying models immediately.

conceptAgentic Commerce Protocol (ACP)

A shared standard developed with OpenAI for businesses to communicate with AI agents, enabling agents to make purchases on behalf of users.

companyDiori

A large Shopify merchant mentioned as coming soon to ACP integration.

companyCrunchAi

A company that does web rewriting for agents, with customers like u-hims and scams.

productLink

Stripe's consumer product, a dense network with over 200 million consumers, showing high adoption in the AI network (e.g., 58% of Lovable's volume).

personKatie Dill

Stripe's Head of Design, known for her uncompromising stance on quality and obsession with every detail of user experience.

productRobin Hood Credit Card

Credit card marketed with 'free trial cards' that are designed to expire after 24 hours to avoid charges for free trials.

companyPhil's Coffee

A coffee shop mentioned in the DoorDash example.

mediaForbes AI50

A list of top AI companies, all of which monetize online through Stripe.

softwareHubert

Stripe's internal AI tool for natural language to SQL queries, designed to help users ask questions about business data.

softwareKronan

The open-source name for Stripe's internally built feature engineering platform, 'Shepard'.

companyTecton

A feature engineering platform that an internal Stripe team wanted to adopt, but was ultimately not used on the charge path due to latency/reliability concerns.

companyGlossier

A large Shopify merchant mentioned as coming soon to ACP integration.

softwareRadar

Stripe's machine learning system designed to block fraud for its customers.

companyShade Form

A YC startup that accepts stablecoins, with 20% of their volume coming from stablecoin payments, demonstrating their utility for global transactions and high price points.

productToken Billing

An API launched by Stripe that allows businesses to track and price services to inference costs in real-time, helping them adjust prices as LLM costs change.

conceptStripe Sessions

An event where Stripe discussed their sandbox product with Replet.

softwareCoursera

productHubble