Updated
Updated · Business Insider · Jun 11
Nadella Urges Microsoft's 220,000 Staff to Curb AI Overuse as Model Costs Come Due
Updated
Updated · Business Insider · Jun 11

Nadella Urges Microsoft's 220,000 Staff to Curb AI Overuse as Model Costs Come Due

3 articles · Updated · Business Insider · Jun 11

Summary

  • Satya Nadella told Microsoft employees not to default to frontier AI models, saying the company should match each task to the right model rather than chase maximum token use.
  • At a live Hard Fork podcast taping, Nadella said “tokenmaxxing” inside Microsoft is widespread and addictive, but argued AI use must deliver both useful output and sound economics.
  • He pointed to Copilot’s auto mode as the practical answer, routing work to the most appropriate model instead of using the most powerful system for routine problems.
  • Nadella also described using AI to maintain a software project by tracking workplace discussions, part of his push to make Microsoft’s 220,000-person workforce move faster in the AI era.
  • The remarks reflect a broader shift in Silicon Valley from encouraging maximal AI usage to scrutinizing the cost of heavy model consumption as companies face rising bills.

Insights

Is Microsoft's push for 'token-efficient' AI a sign the technology's economic bubble is bursting?
As Microsoft builds its own models, is the landmark partnership with OpenAI starting to fracture?
With AI agents monitoring workplace chats, what is the new line between productivity and total surveillance?