Scheduled Tasks

A computer usually does most things only once a human sets them in motion. Scheduled Tasks reverse this relationship. You define once what should happen and when it should happen. After that, the task runs on its own, again and again, without anyone having to think about it. The German term for this is “geplante Aufgaben.” The principle resembles an alarm clock: you set it once to 6:30 a.m., and after that it rings every morning by itself.

Why automation by time is so widespread

Very many tasks in technology are boring but indispensable. Data backups belong to this category, as does deleting old files or installing updates. A human will sooner or later forget such tasks. A machine never forgets them. That is exactly where the practical value lies: reliability without attention.

There is a second reason, which has to do with money. Computing power is often cheaper and less utilized at night than during the day. Large evaluations are therefore deliberately scheduled for the night. In the morning, the result is ready and waiting, even though no one was awake for it.

For AI providers, scheduled tasks are also a step away from the pure question-and-answer game. A chatbot that only reacts is a tool. A system that acts on its own seems more like an assistant. That is why Scheduled Tasks appear in product announcements as an argument for “agents” — that is, programs that independently carry out multiple steps.

From schedule to execution

At its core, three pieces of information are needed: a task, a point in time, and a repetition rule. The task is a command or a program. The point in time is a time of day or a date. The repetition rule states whether the whole thing happens once, daily, weekly, or monthly. In the background, a service runs that constantly checks the clock and starts whatever is due at the right minute.

On Linux servers, this system is traditionally called Cron, and the individual entries are called cron jobs. Windows has a Task Scheduler with a graphical interface. With AI services, on the other hand, you formulate the request in ordinary language, for example “Summarize the most important news on semiconductors for me every Monday at eight.” The model translates this sentence into a schedule.

An often overlooked point is important here: the task runs on a server, not on one’s own device. The phone is therefore allowed to be off. Conversely, this means that errors can go unnoticed. If access expires or a website changes its structure, the task fails silently. Good systems therefore actively report such failures instead of simply delivering nothing.

Between the server room and the chat window

On one’s own computer, scheduled tasks are usually encountered invisibly. The automatic Windows update is one, as is the nightly virus scan or the Time Machine backup on the Mac. The calendar that reminds you of a birthday two days in advance also works according to the same pattern.

In the AI world, the feature has been a visible product characteristic since 2025. ChatGPT offers “Tasks,” with which one can set up recurring assignments. Other providers have followed suit. Typical applications are a daily news overview, weekly monitoring of prices, or a reminder that delivers matching context at the right moment.

In business news, Scheduled Tasks often come up in connection with costs. Every automatic execution consumes computing time, even if no one reads the result. Providers therefore limit the number of tasks per user. And there is a risk of confusion: a schedule does not make a program autonomous. It does not decide when it would make sense to act. It merely adheres to the clock that a human has set.