Microsoft CEO Satya Nadella is promoting a new narrative around artificial intelligence, urging stakeholders to focus on how AI tools empower humans and how resources should be allocated to support this technological expansion.

He’s now exercising “thought leadership,” publishing on a new blog called “sn scratchpad” that he promoted through his company’s LinkedIn platform.

Nadella writes early in his post that “We have moved past the initial phase of discovery and are entering a phase of widespread diffusion.” While Pew Research indicates 62 percent of U.S. adults interact with AI at least several times weekly, the more critical metric for Microsoft remains how many customers actually pay for Copilot and related AI services—a figure that continues to develop.

He also introduces his version of the tech executive deflection phrase “it’s early days,” stating “We are still in the opening miles of a marathon. Much remains unpredictable.”

Nadella outlines three priorities the technology industry and society must address in 2026 to make AI deliver genuine value. First, he argues for developing a “theory of mind” that positions AI as a tool amplifying human capability, with products designed around this principle. He references Steve Jobs’ characterization of computers as “bicycles for the mind” as his foundation.

Given that Microsoft’s own research on AI’s labor impact has required reassurance that it doesn’t predict widespread job elimination, Nadella’s emphasis on AI as human augmentation rather than replacement becomes understandable.

“What matters is not the power of any given model, but how people choose to apply it to achieve their goals,” Nadella stated.

Second, he advocates moving from individual “models” to integrated “systems” where multiple models and agents collaborate to create value. He describes building “rich scaffolds that orchestrate multiple models and agents; account for memory and entitlements; enable rich and safe ‘tools use.'”

Third, Nadella contends society must make strategic decisions about AI deployment to achieve “real world impact.” He writes, “The choices we make about where we apply our scarce energy, compute, and talent resources will matter. This is the socio-technical issue we need to build consensus around.”

Incognito chats aren’t saved to history or used to train models.