How AI Agents Reshape Knowledge Work - Perplexity
Summary
A collaborative study by Perplexity and Harvard Business School researchers, published June 8, 2026, analyzes the impact of Perplexity Computer, a general-purpose AI agent orchestrator launched February 25, 2026, on knowledge work. The findings indicate that Computer significantly expands both the breadth and depth of user capabilities at lower costs. Computer queries grew 84x by May 27, 2026, demonstrating 48x more machine execution time per session (26 minutes vs. 33 seconds for Search) and a 55% reduction in user dissatisfaction. The agent reduces estimated task time by 87% (from 269 to 36 minutes) and task cost by 94% on average, with programming seeing a 96% cost reduction. Furthermore, Computer users engage in cross-occupational tasks 59% of the time and perform more cognitively complex, higher-order tasks, requiring 38% more O*NET Knowledge areas than Search.
Key takeaway
For Directors of AI/ML evaluating agent adoption, Perplexity Computer's empirical results suggest a significant shift in knowledge work. You should consider deploying general-purpose agents to reduce task time by up to 87% and costs by 94%, freeing your teams from manual execution to focus on higher-order, cross-disciplinary tasks. This enables your organization to tackle more complex projects with fewer resources, redefining roles and team structures.
Key insights
AI agents like Perplexity Computer autonomously execute complex workflows, significantly reducing human effort and cost while expanding the scope and cognitive complexity of tasks.
Principles
- AI agents shift user effort from execution to supervision.
- Increased AI autonomy can enhance efficiency and quality.
- Agents enable users to cross disciplinary boundaries.
Method
The study compared Perplexity Search + Human versus Computer + Human workflows, triangulating efficiency via tool-based, LLM-based, and user-reported estimates. Task scope was analyzed using cross-occupation activity and O*NET classifications.
In practice
- Apply agents to long, multi-step workflows.
- Use agents for generative tasks like code or documents.
- Employ agents for work spanning multiple tools.
Topics
- AI Agents
- Perplexity Computer
- Knowledge Work
- Task Automation
- Workflow Efficiency
- Cognitive Complexity
Best for: Executive, AI Product Manager, Product Manager, Research Scientist, Director of AI/ML, Consultant
Related on AIssential
Editorial summary, takeaway, and curation by AIssential. Original article published by perplexity.ai via Google News.