> For the complete documentation index, see [llms.txt](https://docs.ilert.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.ilert.com/on-call-management-and-escalations/on-call-schedules/recurring-schedules/using-ilert-ai-for-schedule-generation.md).
# Create schedules using ilert AI
Describe a rotation in plain language and let ilert AI fill in the recurring schedule form for you to review and save.
Instead of building a recurring schedule layer by layer, describe it in a chat, for example "Weekly rotation with Helena and Humza, handover Monday at 9". ilert AI asks for anything that's missing, then fills in the **Create recurring schedule** form. Nothing is saved until you review the form and click **Save**.
The assistant is in beta. It only works while you create a recurring schedule, not when you edit one or create a static schedule.
## Before you start
The assistant needs AI features to be on for your account. New accounts start with them on. If **Generate schedule with ilert AI** is missing from **Create recurring schedule**, ask your account owner to open **Account settings** and turn on **Allow additional AI features** under **AI features**.
## Generate a schedule
{% stepper %}
{% step %}
### Open the assistant
Go to **On-call schedules** and click **Create schedule**. On the **Recurring schedule** card, click **Continue**. Then click **Generate schedule with ilert AI** at the top right.
A panel opens with the prompt **Please describe the schedule you'd like to create.** The warning at the top means what it says: generating replaces anything you already typed into the form, but saves nothing.
{% endstep %}
{% step %}
### Describe the schedule
Type what you need and send it with the arrow button, or with **⌘ + ↵** on a Mac and **Ctrl + ↵** elsewhere. **↵** on its own starts a new line.
The assistant asks one question at a time, in this order, and skips what you already told it:
1. Who is on call. Name people, teams, or both. The assistant looks up each person in your account and asks you to clarify when a name matches nobody or several people. For a team, it takes the first team whose name matches and sees at most 20 of its members. It adds you only if you say so.
2. When the schedule starts.
3. How often the shift passes to the next person, in days or weeks, and whether the schedule ends. A monthly rotation becomes 4 weeks.
4. Whether coverage is limited to certain hours, such as working hours. Without a limit, the schedule covers 24 hours a day.
It replies in German if your first message is in German, and in English otherwise.
{% endstep %}
{% step %}
### Generate
When the assistant has what it needs, it replies **Thanks! Please click the button below to generate the schedule.** Click **Generate schedule**. The page reloads with the form filled in.
{% endstep %}
{% step %}
### Review and save
Check every field before you click **Save**, especially **Timezone** and the time in **Starts on**, which sets the handover. The assistant uses Europe/Berlin unless you name another timezone, even if your account uses a different one.
The assistant fills in **Name**, **Timezone** and, for each layer, the users, **Rotate every**, **Starts on**, an end date if you gave one, and the coverage. It doesn't set **Teams**.
{% endstep %}
{% endstepper %}
The panel before the first message. The example in the message box shows the kind of request it understands.
The new schedule appears in **On-call schedules**. Like any schedule, it pages nobody until an [escalation policy](/on-call-management-and-escalations/escalation-policies.md) uses it.
## What the assistant can't set
Finish these in the form after generating, before you save:
* **Teams.**
* **Follow support hours (respects holidays)** and holiday exceptions. Limits on hours always come out as **Only on specific times of the week**. See [Follow support hours](/on-call-management-and-escalations/on-call-schedules/recurring-schedules.md#follow-support-hours-respects-holidays).
* Layer names.
* Primary and secondary rotations. They belong in separate schedules, one for each level of your escalation policy.
For anything else about layers, see [Recurring schedules](/on-call-management-and-escalations/on-call-schedules/recurring-schedules.md).
## If something goes wrong
| What you see | What to do |
| ----------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Your request timed out. Try again with the same prompt?** | Click **Yes** to ask the assistant to carry on. It doesn't resend your message, so answer again if it repeats a question. Click **No** to clear the conversation and start over. |
| **Something unexpected happened. Please try again later.** | The same **Yes** and **No** choices. |
| **Was unable to process your request, please try again.** | Send your last message again. |
| The page shows an error after **Generate schedule** | The result contained something the form can't use, most often a person ilert couldn't find. Start again and name people exactly as they appear in ilert. |
Closing the panel keeps the conversation until you reload the page. The assistant only helps with schedules and declines other requests.
## Your data
The panel says: **Do not share any sensitive data.** Keep passwords, keys and personal details that the schedule doesn't need out of your messages. What you type goes to an external large language model, together with what the assistant looks up in your account to match the people and teams you name.
As **Account settings** explains under **AI features**, ilert stores your prompts, the generated output and usage metrics to improve ilert AI. The same section says which model provider handles your prompts, and how to contact support to change it.
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.
Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:
```
GET https://docs.ilert.com/on-call-management-and-escalations/on-call-schedules/recurring-schedules/using-ilert-ai-for-schedule-generation.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
