Schedule, Review and Pause Gab AI Tasks
Schedule a useful request, choose its time zone and delivery, and return to the results. Find Activity, Run Now, editing, pause controls and troubleshooting.
Repeat a request you know how to check
Use Tasks for a recurring brief, reading prompt or report with clear inputs, timing and delivery. Test the request and inspect each result. A saved schedule is a configuration, not a guarantee that every future run will complete.
An actual scheduled brief and a focused rerun
Schedule a recurring Chat task, then adjust its instructions as plans change. This workshop example updates the start time from 09:15 to 10:30 while keeping the other facts.
Facts:
- Cedar workshop
- Saturday at 10:30
- Introductory woodworking
- 12 adult places
Unknowns:
- Venue
- Price
- Registration linkTwo runs, different triggers
| Item | Details |
|---|---|
| Automatic scheduled occurrence | The first workshop brief came from the scheduled occurrence before a Run Now request. Compare its supplied facts and missing details below. |
| Paused schedule | The saved schedule remained paused after returning to Tasks and after the later manual request. |
| Manual revision | The saved request changed 09:15 to 10:30. Run Now returned the revision while preserving the other facts and unknowns. |
Inspect the original scheduled request and result
This output came from the automatic scheduled occurrence; the revision above was a manual rerun.
Using only these facts, return a brief with labels Facts and Unknowns. Facts: Cedar workshop, Saturday, 09:15, introductory woodworking, 12 adult places. Venue, price and registration link are unknown. Do not search, email or publish anything. Keep it under 90 words.Facts
- Workshop: Cedar
- Topic: Introductory woodworking
- Schedule: Saturday, 09:15
- Capacity: 12 adult places
Unknowns
- Venue
- Price
- Registration link1. Set up the request and execution mode
Open Tasks and choose New task or a template. Complete Configure, Schedule, Email and Review. Name the task, supply instructions of up to 2,048 characters and choose Run using and the model.
Choose the execution context
| Item | Details |
|---|---|
| Chat | A standard Chat response using an eligible text model. Standalone tasks and templates default to Chat; each run creates a new conversation. |
| Work | A task with supported tools/files and separate Plus, credit, active-session and launch requirements. A standalone run can create a new Work conversation. |
| An existing Work conversation | A task created in that Work conversation can deliver later runs there and uses that conversation’s main model. Check the binding before saving. |
| Optional bot | Choose an available bot when its reusable instructions fit the task. The bot selection does not grant permissions or replace the task’s changing facts. |
2. Choose cadence, time and time zone
The web picker offers Once, Daily, Weekly and Monthly. Recurring tasks retain their saved time zone when opened on another device; new tasks start with the creating device’s zone. Review cadence, time and zone before saving, especially around daylight-saving changes.
- For Once, check the calendar date and time rather than only the hour.
- For Weekly or Monthly, check the selected weekday or day of month.
- Monthly schedules use local calendar dates. If the chosen day is unavailable, the occurrence uses that month’s final day. Check the intended occurrence for your schedule.
- Review the cadence, time and zone together. If no absolute next-run timestamp is shown, check your calendar and inspect Activity and Results after the expected occurrence.
- Allow time for scheduled processing and model execution after the chosen start time.
3. Review delivery before creating
Choose whether to enable email delivery. With it off, find results through the task’s Results action and conversation history. Review the request, model, schedule and delivery choices, then choose Create Task.
4. Inspect Activity and the actual result
Open the task, choose Activity and read Total Runs, Successful, Failed and Last run. Open View Results or use Results in the list. Compare the actual answer with the source facts; a success count does not prove that the content is correct. Run Now is a manual request and does not establish that the scheduler fired.
Known-answer checks for the example
| Item | Details |
|---|---|
| Supplied facts | Cedar workshop; Saturday; introductory woodworking; 12 adult places. |
| Time change | 09:15 in the automatic result becomes 10:30 in the manual revision. |
| Missing details | Venue, price and registration link remain unknown in both outputs. |
| Observed Activity | After both runs: two total, two successful, zero failed. Only the first run was automatic. |
5. Revise the saved request and test deliberately
Edit the task’s instructions, choose Save changes and inspect the saved text. Name the specific correction and keep other facts stable. Run Now creates another chargeable request; it does not advance the recurring cadence. In the checked example, it also worked while the recurring schedule was paused.
Using only these facts, return a brief with labels Facts and Unknowns. Facts: Cedar workshop, Saturday, 10:30, introductory woodworking, 12 adult places. Venue, price and registration link are unknown. Do not search, email or publish anything. Keep it under 90 words.6. Pause future runs or remove an unwanted schedule
Turn off the task’s Enabled switch and confirm Paused. Reopen Tasks to check the saved state. Pausing future scheduling is separate from stopping work already running, and a deliberate Run Now can still execute. Deleting a result conversation does not delete its schedule. Remove the schedule itself when you intend to stop keeping it.
Review request access and usage
Choose an available model and Chat or Work mode, then review the current estimate. Scheduled runs use applicable credits; Work retains its Plus and task-capacity requirements. Find completed requests in Usage activity.
Task limits and failures
The inspected creation route limits an account to 20 undeleted task records, including paused ones. If creation is refused, review the schedules you intend to keep. Model access, credits, Work capacity and an unavailable bound conversation can also block a run. Inspect Activity and the actual error before retrying; repeated retries may create new usage.
- Check Enabled, the saved local time/zone and whether the expected occurrence is still ahead.
- Check model availability, current account access and credits.
- For Work, inspect active-session capacity and whether a bound task is already busy.
- Review failed-run details. The inspected worker disables a task after three consecutive non-deferred failures; resolve the cause before enabling it again.
- If an associated Work conversation was deleted, the schedule is retained but paused. Review its mode/binding before resuming.
Start with a request you can check
Try a small reading brief or meeting-action request with permitted inputs. Run it once and review the result before choosing a recurring schedule and delivery.
For a weekly report, name the reporting period, source location, measures and output format. Explain what to do when new data is missing, and make sure the scheduled run can access the intended source. For recurring research, ask for source links and changes since the previous period. The task checks at its scheduled time; it is not continuous monitoring.
From these permitted notes: [NOTES], return a short weekly brief with Facts, Actions and Unknowns. Give each action only an explicitly supplied owner/deadline. Do not send messages or publish.
Revision: change only [ONE FACT]; preserve [DETAILS].
Check: compare every row with the notes before using the result.Scheduled task questions
Timing, manual execution and successful content are separate checks.
Can I schedule AI prompts to run every week?
The checked picker supports weekly schedules. Set the weekday and local time, read the saved zone and inspect the first actual run. This walkthrough observed one daily occurrence, not repeated weekly delivery.
Does Run Now prove my schedule is working?
No. It is a manual run. Inspect Activity and a result from the expected scheduled occurrence separately. Run Now does not advance the recurring cadence.
Does pausing prevent a task that is already running?
Pausing controls future scheduled runs. An already-started run has its own state and stop controls; a manual Run Now can also execute while a schedule is paused.
Are all scheduled tasks included at no extra cost?
No. The inspected scheduled Chat calculation adds a 5-credit scheduling surcharge plus applicable model/file/tool usage. Work has its own model/tool and access requirements. Review actual Usage entries rather than treating the sample as a universal price.
Does deleting a result chat delete the schedule?
No. Manage the schedule in Tasks. A schedule attached to a deleted Work conversation is retained but paused; standalone schedules are separate records.
Where do I find the result with email disabled?
Use Results in Tasks or View Results in the task’s Activity, and inspect the generated conversation. Both captured task outputs used this path with email disabled.
Guides and resources
Find more detail and choose your next step.
Schedule a request you want to repeat.
Open Tasks, test a clear request and review the cadence, time zone and delivery before creating the schedule.
