Satya Nadella: the "learning loop," a new competitive advantage for businesses

Microsoft CEO Satya Nadella says business success in the AI economy relies on a learning loop of human capital and token capital, not just top models.

Microsoft's CEO posted a lengthy text on 𝕏 about the future of the company in an AI-driven economy. His central idea: the advantage will not come from choosing the best model, but from an organization's ability to build a "learning loop," a loop where human experience, business processes, and model capabilities continuously reinforce each other.

He distinguishes two types of capital destined to coexist. "Human capital" encompasses the knowledge, judgment, relationships, and pattern recognition specific to teams. "Token capital" refers to the AI capability that the company builds and owns. The former would not lose value as the latter grows, human agency remaining, in his view, the engine of this growth, without which computing power "runs idle."

Nadella advocates for an architecture where each company develops agentic systems that improve with use while retaining control over its intellectual property. The real test, he writes, would be to be able to replace a general-purpose model without losing the "veteran" expertise encoded in its system, through private evaluations and internal reinforcement learning environments. Finally, he warns against a concentration of value on a few models and calls for building a "frontier ecosystem," not just a "frontier model."