Microsoft study: M365 Copilot shifts from search to content work across 5.5M sessions

A new study from Microsoft researchers (Counts et al., arXiv:2605.23958) examines how millions of professionals across over a million companies use M365 Copilot Chat for work. Analyzing approximately 5.5 million anonymized sessions, the team classified user intents and mapped them to O*NET work activities. The findings reveal that writing is the dominant use case, but Copilot is also heavily employed for information retrieval, analysis, decision-making, strategizing, and evaluating/correcting programs and systems. Notably, time trends indicate a relative decline in 'chat as search' behavior and a rise in content creation and communication-related tasks, suggesting users are integrating AI more deeply into core workflows rather than treating it as a simple Q&A tool.

Occupational comparisons show that Copilot usage is broad but uneven: some tasks cut across all job types, while others are occupation-specific. Areas where labor market work is currently underrepresented in Copilot usage point to the next frontier for enterprise AI adoption. The study positions M365 Copilot as an emerging everyday assistant for knowledge work, with its real-time, work-context integration providing a unique window into how generative AI is reshaping professional productivity. Key implications include the need for tailored AI training per job function and the accelerating shift from information retrieval to collaborative content generation.