How AI Autonomy Works
The AI Autonomy module — ticket triage, diagnostic playbooks, the policy engine, and how to configure everything via the dashboard without API calls.
What AI Autonomy does
AI Autonomy is the module that lets artificial intelligence take real actions in your platform — not just suggest things. It classifies tickets, runs diagnostic scripts on devices, and can execute remediations — all under a policy engine that you control. Nothing executes without passing through that gate, and you decide exactly how much autonomy each action type gets.
Plan requirement: Autonomy is an Enterprise plan feature. If you don't see it in the sidebar, contact your account manager.
The three components
1 — AI Ticket Triage
When a new ticket is created, Triage automatically:
- Classifies the issue (Connectivity, Performance, Printer, Software, Storage, Updates, Errors, or Other) with a confidence percentage
- Identifies the affected device from the ticket description or linked device
- Runs the matching diagnostic playbook (if the device is online and the policy mode allows it)
- Posts an internal note tagged [AI Triage] with the classification, device findings, and suggested next steps
This all happens within seconds of ticket creation — before any agent opens the ticket.
2 — Diagnostic Playbooks
Seven built-in read-only playbooks collect live data from agent-based devices:
| Playbook | What it checks |
|---|---|
| Disk Usage | Drive space summary, top 10 largest folders |
| Service Status | Stopped auto-start services, recent crashes |
| Event Log Errors | System and Application errors from last 24h |
| Pending Updates | Pending reboot flags, Windows Update count |
| Network Connectivity | Gateway ping, DNS resolution, public latency |
| Printer Queue | Printer status and queued job count |
| Resource Consumers | Top 10 processes by CPU and memory |
Playbooks can be run manually from the Linked Device panel on any ticket, or run automatically by AI Triage.
3 — The Policy Engine
Every autonomous action — whether from AI Triage, an automation rule, or a manual playbook run — must pass through the Policy Engine. The Policy Engine decides what happens based on the action's configured mode:
| Mode | What happens |
|---|---|
| Disabled | The action never runs; the request is rejected and logged |
| Suggest | The action is queued in Autonomy → Approvals for human review |
| Requires Approval | Same as Suggest — queued, waits for human sign-off |
| Fully Automatic | The action runs immediately within any rate limits you set |
Configuring policies via the dashboard
Go to Autonomy in the sidebar (main page, not Approvals). You will see all 8 known action types as cards — grouped into Diagnostics and Remediations.
For each action:
- Choose a mode — click one of the four mode buttons on the card (Disabled / Suggest only / Requires Approval / Fully Automatic)
- The change saves immediately — no API calls needed
Default behaviour (if you never change anything):
- All 7 read-only diagnostic playbooks default to Fully Automatic — they run immediately when requested
- Remediation actions (restart service, clear temp files, reboot) default to Requires Approval — they queue for human sign-off
The stats panel at the top shows how many actions are configured and how many are in Fully Automatic mode.
The Approvals page
Autonomy → Approvals shows all actions currently queued for human approval. Each row shows:
- The action name (human-readable, not the internal key)
- Risk level — colour-coded: green (read-only), blue (low), yellow (medium), red (high)
- Who or what requested it (AI, automation rule, or a specific user)
- When it was queued
Click View intended action to expand a row and see exactly what the action will do before deciding. Click Approve to let it run, Reject to discard it. Use checkboxes to bulk approve or reject multiple items at once.
Why is the queue empty? If nothing is in the queue, it means either: a) No actions have been requested yet, or b) All configured actions are in Auto mode and run without approval
This is expected and normal. The queue only fills when an action set to Suggest/Approve mode is triggered.
How Autonomy connects to the rest of the platform
Service Desk: AI Triage posts internal notes on tickets. Triage runs are triggered by ticket creation events.
RMM: Diagnostic playbooks execute via the RMM script layer — the same agent and gateway that runs manual scripts.
Monitoring: Device health data is used by Triage to identify which device is likely affected by a new ticket.
Automation: Automation rules can request actions through the Policy Engine — creating a "smart automation" that runs automatically for low-risk actions but queues for approval on high-risk ones.
Notifications: Admins receive in-app notifications whenever a new action is queued for approval.
Where to go next
Was this article helpful?