Spec Sheet 04: Scheduled runs

What your AI can do on a timer, and who approves it, from five vendors' own pages.

Oct 22, 2026

Every AI product now runs work on a timer: a brief built before you sit down, a check-in drafted an hour before the meeting. The vendors' pages differ on what a run can reach, how often it runs, and what happens when a run needs a yes and nobody is there. Spec Sheet 04 puts the five pages side by side, in plain words traced to each vendor's own page. The descriptions are ours; the two short quotes are the vendors'. The desk jobs and the human lock are ours, and the header on each page says which voice you're reading.

Spec Sheet 04 cover: "Scheduled runs." What your AI can do on a timer, and who approves it: five products, in plain words traced to each vendor's own page. A table lists each product, what the vendor calls a scheduled run, how often it runs, and the active cap on the page, checked Oct 21, 2026. Claude: scheduled task; hourly, daily, weekdays, weekly, or on demand; no cap on the page. ChatGPT: scheduled task; hourly on eligible paid plans, once a day on Free, events on Plus, Pro, Business, Enterprise and Edu; 3, 5, 10 or 15 by plan. Gemini: schedule; daily, weekly or monthly in chat, hourly to yearly or on an email in tasks; 10 in chat, 50 in tasks. Grok: automation; once, daily, weekdays, weekly, monthly, yearly, or on an email; no cap on the page. Copilot: scheduled prompt; a time and frequency you set; 10.

Page 2, what each one runs, in plain words traced to the vendors' pages. A scheduled run is a request you save once and your AI runs at a set time, with nobody there to type it. Claude (Anthropic): every paid plan; hourly, daily, weekdays, weekly, or on demand; each run is its own session on Anthropic's servers with the same connectors as any task; a task that needs local files or apps runs only on the computer; the setup form has an approval mode field. ChatGPT (OpenAI): one-time, recurring and monitoring tasks, plus event-triggered tasks on Plus, Pro, Business, Enterprise and Edu; Free runs once a day in a window such as morning, eligible paid plans get hourly runs and exact times; notifications by push, email, or both; a task made inside a project can't read the project's files; active tasks 3 for Free and Go, 5 for Plus, 10 for Business and Edu, 15 for Pro and Enterprise. Gemini (Google): in chat, a schedule prepares a response in the hour before delivery, daily, weekly or monthly, with Keep Activity on; in tasks (AI Pro or Ultra, personal accounts), a schedule triggers a task once, hourly, daily, weekly, monthly, yearly, or on an email or event, with approximate run times. Grok: on grok.com and the apps; once, daily, weekdays, weekly, monthly, yearly, or, with SuperGrok, on an email that matches a filter; instructions, files, connectors and skills attach; every run is a fresh request on current data, saved as a conversation and reported by email, app, both, or neither. Copilot (Microsoft): in Copilot Chat, Teams and Outlook with a Microsoft Copilot license; a time and frequency you set, up to 10 prompts, optional email when ready; responses land in the chat list.

Page 3, when a run asks: who approves a run nobody is watching. Claude: Manual mode pauses and asks before an action; Auto approves reads and screens writes and deletes, and when it blocks one tries a safer way or asks; Skip asks nothing, though deleting a file for good asks in any mode; on Team and Enterprise the Owner decides whether Auto is offered (on by default) and whether write tools can be allowed for all tasks (off by default); the Cowork pages don't say what a run does when it pauses with nobody there. ChatGPT: an action that sends a message or changes outside data may need approval, and the page says "If approval is required, the task pauses until you review it"; workspace admins can control which app actions are allowed. Gemini: designed to ask before sending, changing your data, purchasing, submitting a form, or signing in with Google; a remote-browser sign-in pauses the task until you take over; Google's warning is that a schedule run while you're offline may complete an action you can't stop. Grok's page shows a mode selector and says nothing about approvals; Copilot's describes a prompt rerun and its response with no actions and no approvals. Claude Code, a different surface: a local Claude Code Desktop task in Manual mode "stalls until you approve it" and stays open to answer later; a cloud task has no permission mode and runs writes without asking. Human lock box: a scheduled run reads and drafts; nothing it makes leaves your desk until you've read it; set every write to ask first, and answer the ask yourself, after the read.

Page 4, our routing, what to put on a timer; read-only runs first, and the examples are ours. Runs without you (weekdays 7:00, Thursdays 8:30): the morning brief, built before you sit down from a calendar and inbox the run can only read, with today's meetings, what moved overnight and what's waiting on you; and a check-in draft an hour before the weekly meeting with the calls to make on top; both wait on the page for your read. Thinks with you (Fridays 4:00): a run that reads the week's funder and board threads and lists each open question with the date it was asked, so Monday opens on the list and you write the answers. Builds what you describe (first of the month): a run that turns the program tracker into the two-page update the board sees, same sections every month, saved as a draft that you open, fix and send under your name. A note: every run reads, then saves a draft; before the first run, allow its reads and set anything that writes to ask first; run it once while you watch, answer what it asks, and the next run has what it needs.

The rule to keep is on page 3. ChatGPT's page says a task that needs approval pauses until you review it. Google's says Gemini is designed to ask before it sends, changes your data, purchases, submits a form, or signs in with Google, and that a schedule running while you're offline may complete an action you can't stop. Claude's Cowork pages describe three modes and an approval mode field on the scheduled task, and say nothing about a run that pauses with nobody there; the Claude Code Desktop page says a local task in Manual mode "stalls until you approve it." Grok's and Copilot's pages say nothing about approvals. Under the table sits the human lock: a scheduled run reads and drafts, and nothing it makes leaves your desk until you've read it.

Page 4 is the routing. The morning brief and the Thursday check-in draft run without you, from a calendar and inbox the run can only read. A Friday run that lists the week's open questions with their dates thinks with you. A monthly board update drafted from the program tracker is built from what you describe. Every run reads, then saves a draft; before the first run, allow its reads and set anything that writes to ask first.

Sources, all checked Oct 21, 2026. Anthropic, Claude Help Center: Schedule recurring tasks in Claude Cowork, Use Claude Cowork safely, Get started with Claude Cowork, Use Claude Cowork on Team and Enterprise plans; Claude Code docs: Automate work with scheduled tasks, Schedule recurring tasks in Claude Code Desktop. OpenAI, Help Center: Scheduled tasks in ChatGPT. Google, Gemini Apps Help: Schedules in Gemini Apps, Create & manage schedules for tasks in Gemini, Manage your tasks & workflows in Gemini Apps. Automations in Grok, x.ai. Microsoft Support: Schedule your most used Copilot prompts.

Checked against the sources Oct 21, 2026. Two Google pages disagree on whether a task schedule runs with the device off; both are cited. Caps and plan names are as printed on the day. If a row on this sheet stops matching its source, this post gets a RETIRED line at the top.

Discussion about this post

Ready for more?