What AI may do

Settings for automation are under Settings › AI & automation. By default a person approves every action before it runs.

Automation policy

Per case type and risk level, choose Human in the loop (a person approves each action, from My approvals or the case's Automation tab) or Fully autonomous (low-risk actions run on their own). Every change to the policy is versioned so you can see who changed what.

An action-specific policy shows the action after the case type, for example Incident · reset-password. Saving that card changes only that action; the general Incident policy stays separate. Saving one policy keeps unsaved changes in other cards. Refresh AI policies reloads saved values after asking before discarding drafts.

What the service desk may do on its own

Ask for missing details automatically is off until you switch it on. When on, a new request from a person that leaves at least two of its usual questions unanswered gets one message asking for them, before anyone replies. It never runs for request forms, monitoring or security cases, or after someone on your team has replied, and every message is recorded on the case. The settings page shows how often it helped.

Nothing else happens automatically: priority (beyond your routing rules), replies and resolutions stay with your team.

Connected tools

Under Settings › Integrations, connect outside tools (for example Microsoft 365 or Google Workspace) and manage their encrypted credentials. Credentials are shared by connections to the same provider. Under Settings › AI & automation, review Usage & activity and configure Permissions & policy. Connection settings are managed only in Integrations.

Editing a connection without changing its authentication mode preserves its existing OAuth or API-token setting. If an integration action loses its connection, check execution history before retrying: the remote action may have completed even if its response was lost.