Autonomy & work styles
How much an agent does unattended — set by the account coverage model and CSM work style.
How autonomous is an agent?
An agent never decides its own autonomy — you do, through the coverage model. How much it does unattended follows the account's model, falling back to the assigned CSM's work style:
- Approval mode (high-touch) — *every* action waits for your review before it executes.
- Semi-auto (digital) — internal tasks run automatically, but customer emails queue for your approval.
- Fully automated (pool / tech-touch) — actions execute automatically, and you review the activity log after the fact.
Because it's the *account's* model that governs, a fully-automated tech-touch account and a high-touch enterprise account can be worked by the same agent at completely different autonomy levels.
Where it's set
- Account coverage model — on the account (Coverage) or in Admin Console → Coverage Model.
- CSM work style — per user, in Admin Console → Users & Access.
The guarantee
Whatever the mode, a customer email is never sent without approval in Approval or Semi-auto. Nothing acts silently that your policy hasn't explicitly allowed.
Related: *How Velsano AI agents work* · *The AI Agent Inbox* · *Coverage: assignment and territories*