Scheduled Automations

Scheduled automations run AI on a cadence. They are useful for daily reports, weekly reviews, monthly summaries, recurring sweeps, and timed reminders.

Use scheduled automations for:

• Daily user health reports.
• Weekly pipeline reviews.
• Monthly reporting on a calendar day.
• First- or last-weekday reviews, such as the first Monday or last Friday.
• Morning support digests.
• Weekly marketing ideas.
• Recurring blocker sweeps.

Template prompts can propose scheduled automations as starting points. In the setup conversation, decide which proposals fit the role and confirm the cadence, timezone, destination, and enabled state before any automation is created. The new automation is created disabled unless you explicitly ask for autonomous execution to begin.

Cadence and instructions

Good scheduled automations define:

• How often the automation should run.
• Which sources AI should check.
• What output AI should produce.
• Who should review or receive the result.

Keep the instructions narrow. A scheduled automation should describe the recurring job, while skills should hold reusable workflow details that may apply across many jobs.

Monthly schedules

Monthly automations can run on:

• A numbered calendar day, from the 1st through the 31st.
• The first occurrence of a weekday, such as the first Monday.
• The last occurrence of a weekday, such as the last Friday.

For a numbered day that does not exist in a shorter month, Type runs the automation on that month’s last day. For example, a schedule set for the 31st runs on February’s final day.

Model, reasoning, and speed

When model selection is enabled, scheduled automations in Claude and Codex Spaces can choose how each run executes:

• Agent default is the current UI label for the Space’s default runtime and model.
• A specific model pins every thread created by the automation to that model and its runtime.
• Reasoning choices appear only when the selected model supports them.
• Fast appears only when both the selected model and the automation owner’s provider connection support it. Claude Fast uses Anthropic’s premium speed tier. Codex Fast requires the automation owner’s connected ChatGPT subscription and uses its Fast credits.

The automation owner also funds eligible runs: Type tries that person’s connected Claude or Codex subscription before using workspace credits, even when an integration or teammate triggers the automation.

The Test action uses the same model, reasoning, and speed settings as the scheduled run, including unsaved changes in the editor. This makes the test representative of what the automation will execute.