4 Titans Say AI Agents Are Reshaping Software Economics

For finance professionals, the key signal from Madrona's IA40 Summit isn't just technological—it's economic. If AI agents operate most business software by mid-2027, the pricing power and recurring revenue stability embedded in software valuations come under direct pressure, while new rent shifts to whoever controls the agent layer, customer data, and payment flows.

At Madrona's IA40 Summit on September 30, 2026, at Seattle's Four Seasons, executives from Microsoft, Amazon, Anthropic, and Stripe converged on one of the most destabilizing consensus points in enterprise technology: AI agents are now moving decisively between companies and their customers. The day produced broad agreement that this shift is already underway, but it left hanging what Madrona Managing Director Matt McIlwain called the biggest unresolved question: who ends up owning the customer relationship and the behavioral data generated when an AI assistant acts on a person's behalf. He said versions of that question came up "over and over and over again" during the day, and it maps directly to where profits, pricing power, and brand value will accrue as agentic commerce scales.

“

Microsoft and Amazon have strategic reasons to own the agent orchestration layer, Anthropic's Claude is positioned as a leading computer-use model, and fintech players like Stripe want to own agent commerce.

The context for that consensus is a capability inflection that participants said arrived around July 2026. Jean-Denis Greze, CEO of the agent startup Town, told attendees that since roughly July, AI has been able to operate a browser or a desktop computer almost as well as a human at a reasonable cost. That matters because it means an AI assistant can manipulate an application's user interface directly—clicking, typing, reading screens—rather than relying on a vendor-provided API. The consequence, Greze argued, is that software applications become "thin" shells: the AI does the work, and the underlying app loses its role as the primary unit of software and its grip on usage data. Microsoft EVP Charles Lamanna, who oversees Microsoft 365 and the Copilot platform, made a parallel prediction: most business software will be used by AI assistants on a user's behalf, leaving software makers with less power to set prices. For investors and operators, that is a direct hit to conventional SaaS economics built on per-seat pricing and API-based integration moats.

The implications ripple across distribution, payments, and data flows. Lamanna's "thin app" framing suggests incumbent software vendors may retain some relevance as back-end systems of record, but the customer interface—and the monetization opportunity that comes with it—migrates to whoever manages the agent, the identity layer, or the orchestration platform. Stripe's CRO for AI, Maia Josebachvili, described AI agents conducting commerce on Stripe, which highlights a second-order effect: when agents make purchase decisions, payment providers and digital identity rails become more important, even as merchants risk losing direct consumer relationships. That tension is not academic; it will determine whether SaaS incumbents like Microsoft and Amazon capture new value in the agent stack or watch their core software franchises become cost centers feeding third-party agents.