Perplexity Launches 'Computer' to Automate Workflows
AI startup Perplexity has launched “Computer,” a new system that can autonomously create and execute entire workflows. The platform integrates multiple AI models to orchestrate complex business tasks like data analysis and report generation without human intervention, signaling a major step in automating knowledge work.
Founded in 2022 by Aravind Srinivas, Denis Yarats, Johnny Ho, and Andy Konwinski, Perplexity AI has rapidly achieved a significant valuation. By June 2025, the company's valuation reached a reported $14 billion, with investors including Jeff Bezos and Nvidia. CEO Aravind Srinivas has a background as a research scientist at OpenAI and has held internships at both Google and DeepMind. "Computer" differentiates itself by orchestrating a suite of specialized AI models rather than relying on a single one. At launch, the system integrates 19 different models, using Anthropic's Claude Opus 4.6 for core reasoning, Google's Gemini for in-depth research, and others like xAI's Grok for speed-focused tasks. This multi-model approach allows it to route subtasks to the most suitable AI. The system is designed to move beyond the conversational limits of chatbots by tackling entire workflows that can run for extended periods without direct human oversight. Instead of a user providing a specific prompt for an immediate answer, they describe a desired outcome, and "Computer" breaks it down into a series of autonomous sub-tasks handled by specialized "sub-agents". This new platform is positioned against other emerging AI agentic systems. While tools like AutoGPT provide frameworks for developers, Perplexity Computer is offered as a ready-to-use product. It competes with the workflow automation capabilities being built into systems from OpenAI, Anthropic, and Google's DeepMind. Perplexity Computer is available to subscribers of the company's "Max" plan. The service operates on a usage-based credit system, with Max subscribers receiving a monthly allotment of credits to run these complex, asynchronous tasks.
