OpenAI has introduced a new 'agent mode' for ChatGPT, transforming it into a more autonomous digital assistant capable of performing complex tasks using its own virtual computer.

This new mode allows ChatGPT to 'think' and 'act' independently, enabling it to execute instructions without constant user input, effectively turning it into a digital colleague.

The agent can handle a variety of tasks such as summarizing calendar events, planning meals, creating presentations, and automating weekly reports, making it highly versatile.

It can connect to applications like Gmail, Google Drive, and Github, allowing it to pull information and perform actions across multiple platforms.

The agent is capable of managing personal and professional tasks, including sending emails, analyzing files, creating PowerPoint presentations, and generating detailed reports with citations.

OpenAI envisions this feature revolutionizing productivity and digital task management for users across industries, from students to businesses.

While the capabilities are impressive—such as bilingual presentations and data summarization—there are occasional errors and limitations, prompting calls for improvements.

Users maintain control over the agent, with options to interrupt or take over tasks at any time, though some frustrations include misrecognition of files and autonomous decisions that don't align with user needs.

OpenAI recommends caution, especially for high-stakes tasks involving sensitive data, and plans to enhance control, privacy, and technical stability in future versions.

Activation of the agent mode is straightforward through the Tools menu, with higher usage limits available on Pro plans—up to 500 runs per month—making it accessible for intensive use.

Currently available to Pro users, the agent is expected to roll out to Plus, Team, and eventually Enterprise plans, potentially transforming workflows across various professional settings.

The agent has achieved a new state-of-the-art performance score of 41.6% on expert-level evaluations, indicating significant advancements in AI capabilities.

Announced on July 20, 2025, this feature marks a major step forward in AI development, emphasizing the potential for more autonomous and capable digital assistants.