Automations¶
An automation runs Adverity Atlas on a schedule and tells you only when something needs attention, a spend anomaly, creative that has stopped working, a pacing problem you would otherwise catch too late. It is the difference between asking the same question every Monday and having the answer waiting when it matters.
What an automation is¶
Under the hood an automation is the same agent you use in Chat, given a saved prompt and a trigger. When it runs it investigates, and if it finds something it posts to chat and can notify you.
An automation can also call the tools of an organization-wide MCP server whose allowed consumers include Automation, for example to read a ticket or post a message as part of its run.
A saved prompt: What to investigate, with parameters
A trigger: When to run, a schedule, or an event
Delivery: Post results to chat, and notify on findings
Where automations come from¶
Templates: A library of 31 pre-built automations across budget, pacing, creative, attribution, and governance
Packs: Curated bundles that configure several related templates from one set of shared answers
Custom: Describe what you want monitored in plain language
From a conversation: Turn a chat analysis you already ran into a recurring automation
Triggers¶
Schedule (cron): Run daily, weekly, or on any cron expression, the common case
Event: Run in response to an event rather than a clock
What it produces¶
A run can create artifacts: a report every Monday is just a scheduled automation whose output is a report. When a run finds something, it posts the result to a conversation and notifies the workspace owners. Every run is also recorded as an execution on the Task Executions page, reached from Tasks in the account menu. Automations are independent of artifacts: they can produce them, but neither requires the other.
Note
If an automation fails repeatedly it is paused automatically, so a broken automation does not run unnoticed forever. You are notified, and can fix and re-enable it. This mirrors how webhooks self-disable.
Permissions¶
Roles are cumulative, so the role named is the lowest one that can take the action.
Action |
Required role |
|---|---|
See automations and their runs |
Viewer |
Create from a template or pack, or from a conversation |
Editor |
Create a custom automation from a prompt |
Editor |
Edit an existing automation |
Editor |
Enable or disable, run one now |
Editor |
Stop a run in progress |
The automation’s creator, or Admin |
Delete |
Admin |
What’s next?¶
Creating automations, from a template, from scratch, or from a conversation, and editing one afterwards
Templates, the pre-built library
Tracking runs, history, findings, and cost