OpenAI has officially rolled out a dedicated Scheduled Tasks page in ChatGPT, significantly expanding the platform’s proactive automation and agentic features.
The update centralizes automation management, moving it out of scattered settings submenus and placing it directly into a clean, standalone hub built into the ChatGPT sidebar across web and mobile. Simultaneously, OpenAI announced it is sunsetting Pulse over the next 14 days, effectively replacing it with these highly customizable, user-defined scheduled workflows.
The Scheduled Hub: How It Works
Instead of just reacting to real-time prompts, ChatGPT can now execute background workflows asynchronously while you are offline. The new Scheduled panel functions as a central command station for these automation loops.
- Creating Tasks: You can build a task directly inside the Scheduled tab or naturally in any chat window by simply asking, “Give me an AI tech news briefing every weekday at 8:00 AM” or “Remind me to review my invoices on the first of every month.”
- The Management Panel: From the new page, you can monitor all active automation streams at a glance, check when each prompt is set to trigger next, review historical output logs, or instantly pause, edit, and resume tasks.
- Monitoring Tasks: A major highlight is the introduction of conditional monitoring. You can configure ChatGPT to periodically check an external data source or web page and only trigger a push notification or email when it detects a meaningful change that matches your criteria (e.g., tracking a specific flight price drop).
Active Task Limits By Tier
Because background execution requires consistent server-side compute resources, OpenAI is enforcing strict caps on how many automated workflows can be actively running at any given time based on user subscription levels:
| ChatGPT Plan Tier | Maximum Active Tasks | Primary Deployment Formats |
| Go Users | Up to 3 active tasks | Web and Mobile App (iOS / Android) |
| Plus Users | Up to 5 active tasks | Web and Mobile App (iOS / Android) |
| Business / Edu Users | Up to 10 active tasks | Multi-user workspace arrays |
| Pro / Enterprise Users | Up to 15 active tasks | High-throughput, deep reasoning pipelines |
Operational Mechanics & Safeguards
To prevent loops from hammering system infrastructure, tasks are restricted to running a maximum of once per hour. Furthermore, OpenAI has added an automatic “unattended check” mechanism: if you don’t interact with or click on the outputs of a recurring task for a prolonged period, the system will automatically pause the workflow to conserve bandwidth.
If a user decides to delete an entire chat thread that happens to be tied to a specific background routine, the task will automatically pause rather than breaking completely, allowing you to re-anchor it or safely delete it from the new Scheduled tasks interface without losing contextual formatting.
