Perplexity and Harvard researchers have published the first large-scale study examining real-world AI agent usage, analysing hundreds of millions of queries from Perplexity AI's browser and digital assistant, Comet, which launched in July 2025. The research found that 36 per cent of all tasks assigned to AI agents were productivity and workflow related, whilst 21 per cent involved learning and research, establishing these as the two dominant use cases.
Most AI agent adopters are highly educated individuals from wealthy countries working in technology or knowledge-intensive fields, with the Digital Technology cluster representing 28 per cent of adopters and 30 per cent of queries. Finance professionals dedicate 47 per cent of their queries to productivity tasks, whilst students allocate 43 per cent to learning and research. Users sought more assistance for personal matters than professional ones, with 55 per cent of queries related to after-hours activities compared to 30 per cent for work. Productivity and Workflow categories demonstrate the highest retention rates, as users who initially ask about vacation spots but later use agents to debug Python scripts or summarise financial reports rarely revert to simpler tasks.
This study provides the first empirical evidence of a shift towards a hybrid intelligence economy, where 57 per cent of all agent activity focuses on cognitive work rather than simple administrative tasks. The dominance of cognitive tasks suggests AI agents are fundamentally reshaping how knowledge workers approach productivity and research challenges in both professional and personal contexts, scaling human cognitive capabilities across diverse applications.
Sources:
1. https://www.perplexity.ai/hub/blog/how-people-use-ai-agents
2. https://www.euronews.com/next/2025/12/10/most-people-use-ai-agents-for-productivity-and-learning-perplexity-says
3. https://www.artificialintelligence-news.com/news/perplexity-ai-agents-taking-over-complex-enterprise-tasks/
