Creating automations¶
There are four ways to create an automation, depending on whether a template already fits, you want a whole bundle of related checks at once, you want something bespoke, or you have just run the analysis in Chat and want it to repeat. Once one exists, you can edit it in place rather than starting over.
Prerequisites¶
Editor role, which covers all four routes below and editing afterwards, including a custom automation written from a prompt. At least one data source bound to the workspace, since the automation queries it.
From a template¶
The fastest route. Pick a template, fill in its parameters (which metric, which threshold, which brand), set the schedule, and enable it.
Choose a template: From the library
Set parameters: The specifics for your data
Set the schedule: Daily, weekly, or a custom cron expression
Enable: The automation starts running on schedule
The Create Automation panel asks for a Name, a Trigger (Scheduled (Cron) or Event-based), the Schedule under “When should this automation run?”, and the template’s own Parameters. Between the schedule and the parameters sits Output, which decides what happens when a run finishes:
Option |
What it controls |
Default |
|---|---|---|
Post results to chat |
Whether the run’s result appears as a conversation you can read and follow up in |
On |
Notify me when this automation completes |
Whether a notification goes to whoever created the automation, on every completion, whether or not the run found anything |
On |
Both are tick boxes, and both can be changed later from the automation’s Edit panel, where the second reads Notify the creator when this automation completes, since by then the reader may not be the creator.
Note
A run that fails notifies the creator regardless of this setting, as does an automation paused after repeated failures. The tick box governs completions, not failures.
From a pack¶
When you want a whole area watched rather than one check, use a pack. A pack bundles related templates, asks you for a short set of shared parameters once, and creates every automation in the bundle in one pass. Open Automations > Agent Library, switch to the Packs tab, and click Use Pack.
Packs need the Editor role, the same as creating from a single template.
Custom, from a prompt¶
When no template fits, ask for it in Chat and describe what you want monitored in plain language. Adverity Atlas turns your description into the automation’s prompt. This route lives in the conversation rather than in the automations list, whose Create from Library button covers the template and pack routes only.
Describe the job: “Every Monday, check last week’s spend by channel against plan and flag anything more than 10% off”
Review before it is created: Adverity Atlas shows you the prompt the automation would run, and creates it only once you agree
Set schedule and delivery: The trigger, and whether results post to chat and notify you
Tip
Write a custom prompt the way you would ask a colleague, the condition to watch, the threshold that matters, and what counts as worth flagging. Specific beats vague: “flag CPA above €40” runs better than “watch CPA”.
From a conversation¶
If you have already run an analysis in Chat that you would like to repeat, turn that conversation into an automation. It carries the prompt over, and you add a schedule. This is often the most natural path, you discover a useful question by asking it, then make it recurring.
Editing an existing automation¶
An automation can be changed in place. It keeps its identity, its run history, and its schedule where you left them unchanged, so tightening a threshold does not mean rebuilding and losing every past run.
Open the automation’s detail page and click Edit. A panel opens headed with the automation’s name and, beneath it, either the template it came from or Custom automation.
Adjust what you need:
Field
What it holds
Name
The automation’s name. Required
Description
What the automation does. The field prompts “What does this automation do?”
Instructions
The step-by-step instructions the agent follows. Required, and shown for custom automations only
Trigger
Scheduled (Cron) or Event-based. Required
Schedule
The cron schedule, shown for a scheduled trigger, under “When should this automation run?”
Output
Post results to chat and Notify the creator when this automation completes, the two delivery tick boxes
Parameters
The template’s own parameters, shown for an automation created from a template
Watch the Preview beneath the form. It shows the prompt the agent will actually receive and re-renders as you type, so you see the effect of a change before committing to it.
Click Save Changes. You are returned to the detail page.
Which fields you get depends on where the automation came from. An automation created from the library keeps the library’s prompt and is tuned through its Parameters; only a custom automation has editable Instructions.
Note
An automation that belongs to a pack can be edited on its own, and the panel says so: “This automation is part of a pack. Changes here apply to it alone; its parameters will differ from the pack’s shared values.”
Editing from Chat¶
You can also ask in Chat to change an automation’s schedule, a threshold, or the wording of a custom automation’s instructions. Adverity Atlas first shows you what the automation would run after the change, without saving it, and applies the change only once you approve.
Scheduling¶
Schedules use cron expressions, so anything from “every weekday at 7am” to “the first of each month” is expressible. Event-based triggers run in response to an event instead of a clock, for automations that should react rather than poll.
What’s next?¶
Templates, start from a pre-built automation
Tracking runs, watch what your automation does
Chat, where a from-conversation automation begins