ChatGPT Work turns execution location into a visible product choice because delegated work may need either local context or cloud persistence. This boundary matters now because general chat is being split from longer-running work that creates reusable artifacts and can continue after the conversation window closes.
OpenAI describes ChatGPT Work as a mode for substantial tasks that produce files, analyses, workflows, or recurring updates for human review. The desktop product lets users choose local work for computer-bound files and apps or cloud work for persistence and cross-device continuation. Work can also use files, plugins, and approved tools while exposing progress, questions, direction changes, and approval points.
The product binds a clear outcome to files, plugins, approved tools, progress updates, review checkpoints, and an execution environment chosen for the task's data and duration. Local execution preserves access to computer-bound context; cloud execution preserves continuity when the device is offline and supports scheduled monitoring.
Operators should choose local or cloud execution from the required data boundary, tool access, duration, and approval path instead of treating the modes as cosmetic interfaces. This is a concrete product expression of the agent-ready capability pattern.
The documentation explains the intended workflow but does not establish comparative reliability, privacy, or cost across local and cloud execution for real organizations. Watch whether users can understand where files and actions run, recover interrupted work, and review every important side effect without learning internal infrastructure terms.