Perplexity researchers published empirical findings showing their AI agent product Computer performs 26 minutes of autonomous work per user session, compared to 33 seconds for traditional search. The research, based on production data from real users, quantifies the dramatic shift from search-based to agent-based knowledge work.
Study Uses Natural Experiments From Production Data
The paper, titled "How AI Agents Reshape Knowledge Work: Autonomy, Efficiency, and Scope," was authored by Jeremy Yang, Kate Zyskowski, Noah Yonack, and Jerry Ma from Perplexity and published on arXiv on June 5, 2026. The research analyzed sessions with near-identical initial query pairs as natural experiments, comparing outcomes when users attempted the same underlying task with both Perplexity's Search and Computer products.
Three Key Findings on Autonomy, Efficiency, and Scope
The research identified three primary ways AI agents reshape knowledge work:
Autonomy: Computer automates task decomposition and execution that Search users would manually orchestrate. This autonomy shifts follow-up query distribution toward higher-order work such as verification and extension. Per-query dissatisfaction rates were 55% lower on Computer than Search, indicating improved execution quality.
Efficiency: Computer reduced completion time from 269 to 36 minutes on matched tasks, lowering estimated time by 87% and cost by 94% compared to humans equipped with Search alone.
Scope Expansion: Computer changes the scope of work users attempt. Computer queries more often cross occupational boundaries, require higher-order cognition, draw on broader expertise, take the form of composite tasks that bundle interdependent subtasks, and unlock work activities essentially absent from Search usage among the same users.
Significance for Knowledge Work Automation
The 26 minutes versus 33 seconds comparison represents a 47x increase in autonomous work per session, marking a paradigm shift from information retrieval to automated task execution. According to the paper's abstract, "Together, the evidence indicates that AI agents accelerate workflows, enhance output quality, reduce costs, and expand the breadth and depth of automated work."
This represents significant primary research from a major AI company using real production telemetry rather than synthetic benchmarks, providing empirical evidence for the agent versus search paradigm shift in knowledge work.
Key Takeaways
- Perplexity's Computer performs 26 minutes of autonomous work per session versus 33 seconds for traditional Search
- AI agents reduced task completion time from 269 to 36 minutes, an 87% reduction
- Per-query dissatisfaction rates were 55% lower on Computer than Search products
- Cost savings reached 94% compared to humans using Search alone for matched tasks
- Computer unlocked new categories of work absent from Search usage, expanding the scope of automated knowledge work