AI agents stop where you decide
This is our design decision, not a configuration option. Agents work up to the limit you set, then ask for human approval.
Agents handle repetitive steps, then stop at the approval gate you set. Every action is logged.
A good fit when: tickets and incoming letters are processed by hand.
1. The agent prepares a draft
Reading, summarizing, filling in forms, matching data, flagging anomalies.
2. Approval gate
The agent stops. A person you appoint reviews and approves, with enough information.
3. The action is carried out
Only approved actions are executed in connected systems.
4. Everything is logged
Audit trail: every agent step and every approval can be traced.
Approval points are set together for each workflow: who approves, when, and with what information.