Agent Economy's Impact on Commerce, Fraud, and Monetization

The Agent Economy is Here, and It's Rewriting the Rules of Commerce and Fraud

In this conversation, Emily Glassberg Sands, head of data and AI at Stripe, offers a panoramic view of the seismic shift occurring in the internet economy, moving from a human-centric model to one increasingly dominated by autonomous agents. The implications are profound, revealing hidden consequences for fraud detection, business scaling, and monetization strategies. This analysis is crucial for founders, product managers, and anyone building or operating in the digital space, providing a strategic advantage by illuminating the non-obvious dynamics of this emerging agent-native world. Understanding these shifts is not just about adapting; it's about building the foundational infrastructure for future economic activity.

The New Frontier of Fraud: Compute Theft and the Full-Funnel Battle

The most immediate and perhaps startling revelation from the conversation is the fundamental redefinition of fraud in the age of AI. Historically, fraud revolved around the pilfering of payment credentials or direct financial theft. However, as Glassberg Sands explains, the advent of AI has shifted the target to a more insidious form: compute theft. This isn't about stealing money directly, but about exploiting free trials and credits to consume expensive AI processing power, an existential threat to many AI-native businesses.

This shift necessitates a complete overhaul of how fraud is detected. Stripe Radar, once a transaction-level tool, must now operate across the entire customer lifecycle. The problem begins at sign-up with multi-account abuse, where fraudsters create numerous aliases to claim new user credits. This is not a fringe issue; it accounts for a significant portion of sign-ups for AI companies on Stripe.

The next layer is free trial abuse, a particularly urgent problem where the unit economics break down rapidly. Glassberg Sands highlights a stark example: a company losing $25 per free trial due to LLM spend, with only 4% converting to paid customers. This means a staggering $625 cost per eventual paying customer before any revenue is generated. The proliferation of virtual cards, often marketed for legitimate consumer convenience, exacerbates this, as fraudsters leverage them to bypass payment safeguards. The scale of this abuse is staggering, with one large AI user blocking 250,000 fraudulent free trials weekly, a 4x increase in just six months.

Beyond free trials, non-payment abuse, where customers incur significant overages or fail to pay invoices after consuming services, presents another massive challenge. This moves fraud from a discrete transaction event to a continuous, full-funnel problem.

"The problem with blocking all virtual cards is for AI companies about 15% of legitimate card transactions on Stripe are actually virtual cards... you really don't want to be in the same way you know you don't want to be turning off free trials you don't want to be throttling virtual card virtual cards either."

This comprehensive, full-funnel approach to fraud detection is where the AI arms race truly plays out. While fraudsters leverage AI, so too do the defenders. Stripe's strategy, by processing 2% of global GDP and leveraging its extensive data across various payment methods and processors via the Radar API, provides a crucial advantage. This comprehensive view, treating fraud mitigation as a "public good," allows for disproportionate investment in defenses, enabling them to "eek ahead" in this perpetual battle.

The Unprecedented Growth of AI Companies: Net New Spend and Shifting Business Models

One of the most striking data points from the conversation is the explosive revenue growth of AI companies. Glassberg Sands reveals that the top 100 AI companies on Stripe reach $30 million in Annual Recurring Revenue (ARR) in approximately 18 months, a rate three times faster than the top SaaS companies from 2018. This rapid scaling is attributed primarily to net new spend entering the economy, rather than a direct substitution for existing SaaS or headcount.

This dynamic is fueled by a few factors. Firstly, AI spending is largely experimental, with organizations still learning and often bound by existing contracts. Secondly, AI services, while becoming cheaper, are far from free, representing a genuine new cost center. Glassberg Sands anticipates this will evolve, with AI eventually substituting for traditional SaaS and headcount op-ex. As companies integrate AI into their operations, the cost of an engineer might increase by 10% due to LLM expenses, forcing a re-evaluation of budgets and ROI calculations.

The monetization models are also undergoing rapid iteration. Traditional seat-based pricing, common in SaaS where marginal costs were near zero, is proving inadequate for AI's inherent inference costs. This has led to a surge in usage-based billing, metering tokens, API calls, workflows, and crucially, outcomes.

"Usage based billing has become very important very quickly companies are metering you know tokens and api calls but they're also metering workflows and they're metering outcomes kind of like whatever unit best reflects both the the customer value and the cost structure."

This rapid experimentation with pricing--from subscriptions with overages to prepaid credits and real-time top-ups--reflects the need to align revenue with both customer value and the fluctuating cost of underlying models. The implication is that seat-based licensing, particularly for developer tools, may become obsolete as agents dramatically increase individual developer productivity.

Agentic Commerce: From Assisted Buying to Autonomous Purchasing

The conversation pivots to the profound implications of agents becoming economic actors. Glassberg Sands frames "agentic commerce" not as a singular extreme, but as a spectrum. At its earliest stage, AI simply removes friction from existing online experiences, aiding research and comparison. The next step involves more descriptive search and then delegation, where agents make purchases based on defined constraints. The furthest end of the spectrum is ambient commerce, where systems anticipate needs without explicit prompting.

This shift necessitates a redesign of payments infrastructure. The traditional checkout flow--human at a browser, filling forms, entering card details--is being disrupted. Stripe's "Agentic Commerce Protocol," co-created with OpenAI, provides a shared technical language for AI systems and businesses. This allows merchants to integrate once and then be accessible through various agents and AI interfaces, maintaining their role as the merchant of record while abstracting away the complexity of integrating with numerous new storefronts.

The current volume of agentic commerce is relatively small but growing rapidly, particularly for commodities like Halloween costumes. This mirrors the early days of e-commerce, where higher-priced or quality-dependent purchases were initially avoided. Stripe's consumer wallet, Link, is evolving to support this delegated authority model, allowing consumers to grant specific agents controlled access to their payment credentials with defined guardrails and approval limits.

"The merchant remains the merchant of record and that part really matters like businesses want access to these new storefronts these new channels but they don't want to give up the customer relationship they don't want to give up control over trust or fraud."

This evolution presents a clear advantage for businesses that can adapt their infrastructure to support these new transactional paradigms, ensuring they can participate in these emerging agent-driven marketplaces without sacrificing control over customer relationships or security.

Key Action Items

• Immediate Action (0-3 Months):

  • Assess Current Fraud Exposure: For AI-native companies, integrate Stripe Radar at the sign-up stage to gain visibility into multi-account abuse and free trial abuse. For all businesses, review existing fraud dashboards and consider using the Radar Assistant to understand specific risks.
  • Evaluate Monetization Models: Analyze current pricing structures. If using traditional seat-based licenses, especially for developer tools, begin exploring usage-based or outcome-based alternatives.
  • Explore Agentic Commerce Capabilities: For merchants, investigate how to integrate with the Agentic Commerce Protocol to expose product catalogs and checkout flows to AI agents, even if starting with a toggle-on approach.
  • Investigate Stripe Projects: If you are a developer or builder, explore Stripe Projects for a more streamlined command-line experience for provisioning and managing your software stack.

• Medium-Term Investment (3-12 Months):

  • Implement Full-Funnel Fraud Detection: For companies experiencing rapid growth, especially in AI, fully adopt a full-funnel fraud detection strategy that covers sign-up, free trials, overages, and payments. This is a critical investment to prevent existential threats.
  • Develop Outcome-Based Pricing: Begin the process of defining and measuring key outcomes for your product or service. This will be essential for aligning customer value with your pricing as the market matures beyond token-based models.
  • Prepare for Agent-Facilitated Transactions: For businesses, start to understand the technical requirements and benefits of supporting agent-assisted buying experiences, focusing on secure credential passing and fraud scoring.

• Longer-Term Strategic Investment (12-18+ Months):

  • Rethink Developer Experience for Agents: As agents become builders, consider how your developer tools and documentation need to evolve to support machine interactions, not just human ones.
  • Embrace Delegated Authority for Consumers: For companies serving consumers, explore how to leverage secure delegated authority models (like Stripe Link's evolution) to enable safe agent-driven purchases, fostering trust and reducing friction.
  • Adapt to Shifting Headcount Economics: As AI tools become more integrated, anticipate that headcount operational expenses will change. Businesses should begin planning for how AI spend will impact budget allocations and strategic hiring decisions.
  • Build for the Spectrum of Agentic Commerce: Recognize that agentic commerce is a spectrum. Develop flexible infrastructure that can support everything from agent-assisted discovery to fully autonomous purchasing, ensuring your business is ready for future iterations.