Microsoft CEO Satya Nadella on AI's Business Revolution: What Happens to SaaS, OpenAI, and Microsoft? | LIVE from Davos

The session is introduced as an impromptu fireside chat at Davos with Satya Nadella, described as Microsoft's third CEO, born in India and arriving in the United States right after college, together with David Sacks, introduced as the czar of AI and crypto. The first story is about immigration. The speaker and his wife went to college together in India. He came to the US for graduate school, and they married after he had obtained a green card. Because of that sequence, she could not join him.

The workaround was to give up the green card and take an H-1 visa so that his wife could come. He recalls asking at the American embassy in Delhi where the line was to give up a green card, and being told there was no such line because doing so would be a crazy thing in the 90s. The story is offered as an example of how strange US immigration policy could be, and it is described as a long-lost memory that worked out in the end.

The conversation moves to Copilot, which Microsoft launched first with GitHub and then put on the Windows desktop. The questioner calls the Windows move bold, notes that it came before the assistant could really recognize the file system and interact with applications, and says the reception was lukewarm, though Microsoft has kept doubling down. The questioner sees three modalities emerging for knowledge workers: what xAI is reportedly calling a "human emulator", described as building employees and placing them into chat rooms and email; Claude's newly released co-work, which the questioner says they had been using for 40 hours and found truly impressive; and Microsoft's own approach. The question is how knowledge workers will move from getting interesting results in a chatbot to getting business results.

The answer uses coding as the most illustrative case, calling it probably the best example of knowledge work. The path ran from next-edit suggestions to chat, then to action, and now to fully autonomous agents, which can run in the foreground or background, in the cloud or locally. The speaker says his own belief in this generation of technology took shape when next-edit suggestions started working with real accuracy on an early Codex model, which he places before GPT-3.5.

A key lesson from coding is that developers use all of these form factors rather than picking one. From a command-line interface, a developer can run a foreground agent and a background agent and edit directly in VS Code, all in parallel. Applied to knowledge work, Copilot started with chat. Chat with reasoning goes beyond request and response because users can watch the chain of thought. Copilot now adds actions, through computer use or through skills that an agent calls.

The discussion argues that the AI era needs a new metaphor for how people use computers. It recalls Steve Jobs's line that the computer is a bicycle for the mind, and Bill Gates's line about information at your fingertips. The preferred new metaphor is credited to the CEO of Notion, a product the questioner praises and which Microsoft has not bought.

basically a manager of infinite minds.