KilatBot Articles

Articles

How to schedule AI agent tasks throughout the day

· · Updated

You can schedule KilatBot agent tasks in chat or under Scheduled tasks, then run, pause, resume, or delete them. A sleeping agent wakes when a task is due. For work throughout the day, define inputs, time and timezone, and how to check the result. A schedule is not a guarantee of on-time completion.

Choose work that can stand alone.

Start with one small task that does not need your answer halfway through. Examples to test: summarising an available file, preparing draft captions from a brief, or listing questions for an afternoon review.

Specify what must be ready before the task runs. If a report needs a fresh CSV, decide who prepares it. Scheduling a task does not mean new files are automatically collected from your business sources. Do not add dependencies you have never tested.

Define the output you want to review. For reports, request the period and source. For captions, request a draft label and unanswered questions. If material is missing, instruct the agent to report the gap rather than silently use old material.

How do you write a schedule time?

Schedule examples include every 2h, every day at 07:00, and every monday at 9am. Cron expressions are also supported. Read the Scheduled tasks guide for current steps and controls; this article focuses on planning the work.

Name your business timezone alongside the hour. Example: “every day at 07:00 WIB”. Do not assume your laptop, current location, and agent all use the same time. Before relying on a schedule, compare an actual test run with the time you intended.

An agent made from a template starts on Jakarta time (WIB); a blank agent starts on UTC. Check the timezone in Settings: the times you write in a schedule are read in that timezone. Check one real run before enabling routine work.

If your team works across locations, put the timezone in the task notes, not only the opening conversation. Use one agreed reference when reviewing results. Changed working hours require a schedule review, not an assumption that the old settings remain appropriate.

Start with paused schedules.

KilatBot template schedules start paused. Resume only those you intend to use. Read the task first, especially its data source and intended time. Do not enable every schedule simply because it comes with the template. See how schedules work with memory and tools in the agent overview.

Example request to test:

Prepare a daily summary from the file I specify. My intended schedule is 07:00 WIB. Include the period and filename in the result. If the file is unavailable, report what is missing. Do not message customers or publish the result. Explain the schedule before I rely on it.

Run the task manually first to review its output. Then observe one scheduled execution. The manual test checks whether the work makes sense; the scheduled test checks timing and input readiness under actual conditions.

Does a sleeping agent still run its schedule?

A sleeping agent wakes when a task is due. That does not mean the result finishes at the same instant. Do not count on a report being ready for a meeting based only on its start time. Leave time to read it and handle unfinished work.

Agent memory and files survive sleep and restarts. Storage is not a backup. Keep your own important material.

A scheduled task has nobody present to answer an approval request. Risky commands in such tasks are blocked. Do not design recurring work around automatic approval. Separate preparation from actions needing your decision.

How often can a task run?

The frequency limit is once every five minutes. That is a limit, not a recommendation to run work as often as possible. Choose timing around business needs and input readiness. Avoid two tasks requesting the same report without different purposes.

If results are wrong, pause the task before changing instructions. Check its material and timing, then repeat a small test. Do not assume retries are safe for actions that change data.

For sample inputs, read the daily CSV report article. For weekly drafts, read the content planning guide. Measure your review time before adding tasks throughout the day.

Prepare one test schedule.

Define the task, inputs, timezone, and an outcome you can assess. Start small. Get started.

Read next: An AI daily report generator that reads your CSV file · An AI social media content planner for your week

Built to work for you

Put your agent to work.

Start from a template and shape it to your business.

Get started