AI & Orin
Autonomy Levels (L0–L4)
You control how much the Orin assistant does on its own. Autonomy levels let you adopt AI gradually — start cautious and increase trust over time.
The five levels
Manual Only
AI is disabled for this task. You do everything yourself.
Best for: Highly sensitive processes, or while you evaluate AI.
Suggestion
Orin recommends actions and drafts content, but you make every decision and click to apply.
Best for: A safe starting point for most teams.
Supervised
Orin plans and prepares the action; you review and approve before it executes.
Best for: Once you trust Orin’s suggestions for a task.
Delegated (approval-gated today)
Orin prepares routine actions for one-tap execution and flags anything unusual. In the current release every AI-initiated action still requires your explicit approval before it runs.
Best for: Well-understood, repetitive workflows.
Bounded Autopilot (disabled)
Reserved for pre-approved, low-risk actions inside strict guardrails. This level is disabled in production today — no AI action executes without human approval.
Best for: Future: mature, low-risk processes after evaluation evidence.
How to choose a level
Set autonomy per module (and, in some cases, per task), so you can be cautious where it matters and faster where it doesn't. A common pattern:
- Start at L1 for everything while your team learns the system.
- Move to L2 for tasks where Orin's suggestions are consistently good.
- Consider L3–L4 only for repetitive, low-risk workflows with strong guardrails in place.
Higher autonomy needs stronger guardrails
Changing the level
An administrator sets autonomy levels in Settings. Changes take effect immediately and are themselves recorded in the audit trail.
You're always in control