ChatGPT Work Brings Desktop Automation, Memory and Governance Into the AI Workflow

OpenAI has introduced ChatGPT Work, a desktop-oriented version of its ChatGPT AI tool that aims to bring automation, memory and governance capabilities into AI-powered workflows. This cross-platform environment allows ChatGPT to interact with local applications, files and browser content, potentially enabling tasks such as clicking, typing and moving files within a project. The desktop app combines Chat, Work and Codex functionalities, and its built-in browser supports contextual, end-to-end work.

Key differentiators of ChatGPT Work include its ability to act locally in the background, interact with applications, files and browser content, and utilize plugins, workflows and Scheduled Tasks to automate recurring actions. The tool serves as an agentic layer for work across apps and files, but human approval is still required for certain tasks due to varying permissions and consequences.

Memory in ChatGPT Work allows the AI to retain relevant context across interactions, enhancing personalization when permitted by user settings. This can lead to less repetitive prompting and more connected workflows when relevant work context is available. However, an agent's ability to act on a computer introduces higher permission stakes and requires robust governance practices.

OpenAI provides enterprise governance features, including administrative controls and compliance visibility through the Compliance API, to help organizations manage access, connected tools and action capabilities.

Desktop agents in ChatGPT Work are currently available for Windows and macOS, with similar features also described for web and mobile platforms. OpenAI encourages businesses to carefully map connected applications, identify sensitive files, define approval points and decide which employee roles require access before implementing this technology.