Put Your AI Agent to Work: Schedule and Ship
Run your AI agent on a schedule or trigger in ChatGPT, Claude or Copilot, with a stop plan, run log and hand-off note. 8 lessons, first 2 lessons free.
You’ve built an agent, given it a skill and a connected app, and tested it on your own work. It still waits for you: every Friday, you open it and type the same request. This course moves it to a schedule or a trigger, in the AI tool you already use: ChatGPT, Claude or Microsoft Copilot (formerly Microsoft 365 Copilot) with a work or school account.
You start by choosing the kind of autonomy your job needs and checking your plan offers it. Then you schedule your first run from a self-contained run prompt, add a trigger that drafts and never sends, and write a stop plan you test once. You review real runs and their cost, hand the agent to a colleague with a note that says what it reaches and what it can’t do, plan its upkeep, and check it against twelve acceptance criteria.
Every product step follows the vendors’ own help pages, read on 13 and 14 November 2026, with a route for a plan that can’t schedule. You finish with a shipped agent, Agent Card v4 and the evidence behind it. Lessons 1 and 2 are free.
What You'll Learn
- Evaluate which kind of autonomy a job needs (on demand, a schedule, an event or always-on) and whether your tool and plan offer it
- Apply a schedule to your agent from a self-contained run prompt, and find each run in your tool's task list
- Apply an event trigger, or a scheduled check where your tool has none, with a narrow condition and instructions that draft and never send
- Create a stop plan with run limits, failure thresholds and pause and delete paths, and test the off switch once
- Analyze each run against your Done when line, and track usage so quality and cost stay visible
- Evaluate your agent for shipping: share it with the right rights and a hand-off note, plan its upkeep, and check it against twelve acceptance criteria
After This Course, You Can
What You'll Build
Course Syllabus
Prerequisites
- A tested agent from the course Test and Safeguard Your AI Agent, or any agent with a test set and an approval map
- An account in ChatGPT or Claude, or in Copilot with a work or school Microsoft account
- A recurring job your agent already does well when you start it yourself
Who Is This For?
- Owners of an agent tested in the previous course, ready to let it run without a fresh prompt
- Anyone with a ChatGPT or Claude project, or a Copilot agent, that does the same job every week
- Team leads who want an agent's recurring work reviewed, logged and shareable before others rely on it
- Professionals who want to show a working agent, not a demo
Frequently Asked Questions
Do I need to code?
No. You schedule a run by describing it in plain words, set triggers and approvals with your tool's own switches, and write your stop plan, run log and hand-off note as plain text. The course explains each new term, such as run prompt or stop plan, the first time it uses it.
Which AI tools does it cover, and which plan do I need?
ChatGPT, Claude and Microsoft Copilot with a work or school Microsoft account. Running on its own depends on the plan. OpenAI's help page says ChatGPT Free can keep up to three scheduled tasks that run at most once a day, though its pricing table doesn't tick Free, and that event-triggered tasks need Plus, Pro, Business, Enterprise or Edu. Anthropic puts Claude's scheduled tasks on paid plans. In Copilot, schedules and event tasks run in Cowork, which needs a Copilot licence, usage-based billing and an administrator who turns it on. Where your plan has none of these, each lesson gives a route where you start each run yourself.
Is it safe to let my agent run when I'm not watching?
Only for jobs whose mistakes are easy to catch and undo. Anthropic's own advice for scheduled tasks is to begin with low-risk work such as summaries, and not to schedule tasks that send messages on your behalf, make purchases or take other actions that are hard to undo. The course keeps your agent drafting, with writes blocked or waiting for your approval, and gives it a stop plan and a run log.
Will my agent run all the time once it's scheduled?
No. A scheduled task runs at the times you set, and an event task runs when its event arrives. OpenAI's always-on agent, the dot, is a separate product that OpenAI limits by plan and region, and this course mentions it without teaching it.
Can my colleagues use my agent once it runs on its own?
Yes, in a way that depends on your tool and plan. OpenAI says Free, Go, Plus and Pro users can share a scheduled task as a link; colleagues make their own copy with their own apps, and anyone who opens the link can read its instructions. Claude shares projects on Team and Enterprise plans; on Pro, Max or Ultra you hand over a copy. In Copilot you share your Agent Builder agent with Can chat or Can edit rights. Lesson 6 covers each, with a hand-off note.
Is this course free?
Lessons 1 and 2 are free with a free account. The rest of the course and the certificate come with Pro.
Will I get a certificate?
Yes, with Pro: finish every lesson and score at least 70% on the quizzes for a verifiable certificate.
How long does it take?
Eight lessons; the course page shows the total reading time. Some lessons also ask you to wait for real runs of your agent, which take as long as the schedule you set.
