OpenAI Launches ChatGPT Agent to Handle Real-World Tasks for You
On July 17, 2025, OpenAI debuted its new ChatGPT Agent, a powerful AI tool that merges features from its previous “Operator” and “Deep Research” tools. This agent can autonomously perform complex, multi-step tasks using a virtual computer—browsing websites, interacting with apps, running code, managing emails/calendars, drafting presentations, ordering groceries, and more Cinco Días+15The Verge+15Hindustan Times+15.
In a demo with The Verge, it planned a date by syncing with Google Calendar and booking via OpenTable, generated competitor slide decks, and even ordered office parking—all under user control. Before any irreversible action like purchases or bookings, it prompts for permission The Verge+1Hindustan Times+1. Though task execution can take 15–30 minutes, OpenAI emphasizes the efficiency this offers over manual effort .
OpenAI also integrated robust safety measures: manual approvals, a “Watch Mode” that stops the agent on financial sites, and elevated biological/chemical use safeguards SmartCompany+2The Verge+2Hindustan Times+2. Initially available to Pro, Plus, and Team users, it will soon expand to Enterprise and Education tiers, with global availability following later zh.wikipedia.org+11The Verge+11Wall Street Journal+11.
This launch marks OpenAI’s bold step into the growing AI agents space—competing with offerings from Meta, Google, Anthropic, and others—by moving beyond chat responses to fully integrated task execution
-
It represents a major leap in functionality, turning ChatGPT into a hands-off, task-performing assistant.
-
Offers flexibility and real-world utility, from professional workflows to personal errands—all under user oversight.
-
Puts OpenAI ahead in the AI agent race, taking on industry giants with real-world automation.
