AI Routines - Omni Docs

Routines let you schedule AI-powered tasks to run automatically on a recurring basis. Set up a routine once, and Omni runs your prompt through the Omni Agent on the schedule you define and then delivers the result to the people you choose. Unlike a standard schedule that delivers the same static dashboard or file every time, a routine runs a prompt on each execution. The result is a written, plain-language response — a summary, an answer, or an interpretation of what changed — rather than just a refreshed asset.

Use cases

Routines work best for recurring questions where an AI-written response adds more than a static export would. For example:

Morning briefings for sales teams — “Summarize yesterday’s closed-won deals by rep and flag any account that slipped to next quarter,” delivered before the team logs on
Operational summaries — “Describe how today’s order volume and fulfillment times compare to the same day last week, and call out anything unusual”
Pipeline and revenue narratives — “Write a short update on how we’re tracking against this month’s revenue target and which segments are driving the change”
Plain-language monitoring — “Check for any spikes in error rates or refunds over the last 24 hours and explain what you find”

How routines work

A routine runs your prompt through the Omni Agent on the schedule you set. Each run executes independently and respects your organization’s permissions and data access controls. The result is delivered as text to the destination you choose — by email, or to a Slack channel or users. Each routine runs with the same permissions as the user who created it, so a routine can only query data its creator has access to.

Creating a routine

Restricted Querier permissions or higher are required to create routines.

Routines can be created in the Omni UI, through the API, or from the Omni CLI.

First step

In your Omni instance, click Routines in the side navigation.
Click New routine.
Fill in the fields:
- Routine name — A descriptive name for the routine
- Description — Optional details about what the routine does
- Prompt — The instruction the Omni Agent runs on each execution
- Model — The data model the routine queries, with an optional Topic to focus query generation
- Destination — Where each run is delivered. Choose Email and pick the users or user groups to notify, or Slack and pick a channel or users. Slack requires a connected Slack workspace; to deliver to a private channel, invite the Omni Slackbot to it first.

Define the schedule

In this step, you’ll define the cadence for the routine. The routine’s schedule can be defined using the visual options or with cron:

Visual schedule builder

Easy point-and-click schedule builder

Use the UI options (Daily, Weekly, etc.) to select a time period.By default, schedules are set to send in the local timezone of the delivery creator’s computer. Use the Times are in drop down to change the timezone.

Custom cron schedule (Advanced)

Code-based granular timing control

A cron expression is a string that describes the individual details of a schedule:

Order	Unit	Allowed values	Allowed special characters
1	minute	0-59	* , - /
2	hour	0-23	* , - /
3	day of month	1-31	* , - / L W ?
4	month	1-12 or JAN-DEC	* , - /
5	day of week	1-7 or SUN-SAT	* , - / L W ?
6	year	any	* , - /

Using cron, you can create schedules like the following:

At 9:00 AM every day

0 9 ? * * *

At 6:30AM on the last day of the month

30 6 L * ? *

At 8:45 AM every day, Monday through Friday

45 8 ? * MON-FRI *

Omni uses Amazon Web Services’ (AWS) syntax for cron expressions. Refer to the AWS documentation for more information. By default, the most frequent you can configure a schedule is hourly.

Click the icon above the Cron string field to generate a cron schedule from a prompt like “every weekday at 9AM” or “first Monday of each month at noon”.

Save the routine

When finished, click Save routine.

Use the Create an AI Routine API to create a new routine. The routine will be run with the same permissions as the authenticating user.For example:

curl --request POST \
  --url https://<your-omni-instance>.omniapp.co/api/v1/ai/routines \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "destination": {
    "type": "email",
    "recipientEmails": [
      "blob.ross@blobsrus.com",
      "blob.the.builder@blobsrus.com"
    ],
    "userGroupIds": [
      "550e8400-e29b-41d4-a716-446655440000"
    ]
  },
  "modelId": "3c90c3cc-0d44-4b50-8888-8dd25736052a",
  "name": "Blob Sales morning sales brief",
  "prompt": "Summarize yesterday’s closed-won deals by rep and flag any account that slipped to next quarter,",
  "schedule": "0 9 ? * * *",
  "timezone": "America/New_York",
  "branchId": "3c90c3cc-0d44-4b50-8888-8dd25736052a",
  "description": "Overview of yesterday's deals, including flags for slipped accounts.",
  "topicName": "sales"
}
'

The Omni CLI exposes the AI Routines API as the omni ai-routines command group, so you can create and manage routines from your terminal or a CI/CD pipeline. The request body is passed with --body (or piped in on stdin); run omni ai-routines routine-create --schema to print the body shape.

# Deliver to a Slack channel
omni ai-routines routine-create --body '
{
  "destination": {
    "type": "slack",
    "recipientId": "C01234567",
    "slackRecipientType": "channel"
  },
  "modelId": "3c90c3cc-0d44-4b50-8888-8dd25736052a",
  "name": "Blob Sales morning sales brief",
  "prompt": "Summarize yesterday's closed-won deals by rep and flag any account that slipped to next quarter.",
  "schedule": "0 9 ? * * *",
  "timezone": "America/New_York"
}
'

Other commands in the group mirror the API:

Command	Description
`routines-list`	List routines
`routine-create`	Create a routine
`routine-get`	Get a routine
`routine-update`	Update a routine
`routine-delete`	Delete a routine
`routine-trigger`	Run a routine now

Run omni ai-routines --help to see every command, or omni ai-routines <command> --help for a command’s arguments and flags. See the CLI commands reference for more.

Managing routines

From the Routines page, you can:

View your routines and whether each is active or paused
See when each routine last ran and when it will run next
Edit a routine’s settings, prompt, schedule, or recipients
Trigger an immediate run of the routine
Pause or resume a routine
Delete routines you no longer need

Routines can also be managed with the API or the Omni CLI (omni ai-routines).

Best practices

Be specific in your prompts — Clear, detailed instructions produce more consistent results across automated runs
Test before scheduling — Run your prompt in the Omni Agent first to verify it produces the result you expect before creating a routine
Choose appropriate schedules — Match the routine’s frequency to how often the underlying data changes and when the result is useful
Review results regularly — Check routine output periodically to make sure it still meets your needs as your data and questions evolve

Limitations

Using routines in embedded context is not currently supported
Each routine delivers to a single destination. To send a run to both email and Slack, create one routine per destination.

​Use cases

​How routines work

​Creating a routine

​Managing routines

​Best practices

​Limitations

Use cases

How routines work

Creating a routine

Managing routines

Best practices

Limitations