Automation Alerts are how AI Agent Teams and AI Assistant flows report findings and completed actions back to you. You get a persistent record of what the assistant did or noticed — with supporting data — so you can review, follow up, and archive when resolved.
Where to find alerts
Click the Automation Alerts icon in the sidebar. The panel shows active alerts by default; toggle to view archived alerts.
What appears in an alert
Title and summary — a human-readable description of what the agent found or did.
Source — which connected agent, flow, or action produced the alert.
Attachments — any supporting files. CSV attachments show an inline preview (first 2 rows) with row count, and you can download the full file.
Recurrence — if the alert came from a scheduled flow, the cadence is shown.
Context — the underlying command, API method, parameters, and response provenance are available for debugging or audit.
What generates alerts
AI Agent Teams — Work Order and Equipment Agent Teams post alerts after each scheduled run (e.g., slipping analysis, callback analysis, scheduler runs, data enrichment summaries).
AI Assistant flows — custom flows you configure to run on a schedule or trigger.
Background copilot actions — long-running actions queued from the assistant may surface their final result as an alert.
Managing alerts
Review — read the narrative finding and download attached exports (often Excel files with supporting data).
Archive — archive alerts once you've acted on them. Archived alerts remain searchable.
Toggle view — switch between active and archived alerts from the panel header.
Tips
Alerts are scoped to your user — each person sees the alerts produced for the flows and agents they have access to.
If an alert looks wrong or surprising, check the source agent's flow configuration (schedule, lookback window, custom rules). See How to Configure Agent Flow Schedules and Lookback Windows.
Agents that perform batch actions (like the Work Order Scheduler) post an alert summarizing how many items were touched in the run, so you can spot-check the work.