Key Takeaways

  • "Meta work" is the new mandate: Satya Nadella urges organizations, including Microsoft, to stop merely doing tasks and instead focus on building agentic AI systems that do the work for them. It’s about building the builder.
  • Impossible, not just easier: The ambition should be to achieve what was previously impossible, not simply to make hard tasks easier. This means fundamentally rethinking operational scale and capabilities.
  • Azure's 'Miles' system: Microsoft's Azure networking team exemplified this by creating 'Miles,' an agentic system that manages over 500 fiber operators. Their goal shifted from managing the network to building the AI system that manages it.
  • Tokens over headcount: Nadella proposes measuring operational capacity in terms of AI 'tokens' (processing power for agentic systems) rather than human 'headcount.' This reflects a shift in how large-scale operations are managed and scaled.

The “Meta Work” Mandate: Build the Agent That Does the Work

Most founders are wired to optimize. They identify a problem, then build a solution to do that work better, faster, cheaper. But Microsoft CEO Satya Nadella challenges that entire mindset. He believes the real ambition for builders in the AI era is not to optimize existing work, but to perform "meta work" – that is, to build the agentic system that does the work. Forget making hard things easier; Nadella wants to make the impossible possible.

He uses a striking example from Microsoft's own backyard: the Azure networking team. “They built Essentially, they said, 'Our job is not to do Azure networking. Our job is to build the agentic system that does Azure networking,'” Nadella explained. This wasn't some abstract thought exercise. This was the team managing “the 500 plus fiber operators managing the van, right? All over...” They built a tangible, autonomous AI system they even named 'Miles,' which now handles much of this operational complexity.

The shift is profound. It moves engineering teams from direct operational execution to architecting intelligent, self-managing systems. It’s a move from solving a specific problem to building a general-purpose problem-solver within a defined domain. This isn't just about efficiency; it's about unlocking capabilities that human teams, no matter how large, could never achieve.

From Headcount to Tokens: A New Measure of Scale

This "meta work" philosophy also changes how you think about growth and scale. For decades, scaling an operation meant adding headcount. More demand, more people. But with agentic AI, Nadella envisions a different metric.

Discussing the Azure team's success with 'Miles,' Nadella quoted them saying, “'Look, we don't need head count. We need tokens in order to be able to manage our operation.'” This single line cuts to the core of the change. 'Tokens' here refer to the computational units, the AI processing power, required to run and scale these agentic systems. It suggests that future operational scaling might be measured by available AI capacity, not by human staff.

This isn't about replacing people, but about augmenting their capabilities to an unprecedented degree. It's about designing systems that can grow and adapt autonomously, pushing beyond human-imposed limits. Nadella sees this as the path to truly “make the impossible possible,” giving organizations a new lens through which to view their biggest challenges and ambitions.

What to Do With This

Pick your highest-friction, repeatable operational process that currently consumes significant human effort or is bottlenecked by human limitations. Instead of optimizing the existing steps, map out how an autonomous AI agent would handle it end-to-end. Ask yourself: "What if our goal wasn't to do X, but to build the system that does X?" Then, sketch the first three actions you'd take this week to begin building that agentic system.