✅ Approval vs. autonomous authority

Two independent checks

Before any AI employee uses a tool, Sofilic checks two separate things. Permission: is this tool on this employee's allowed list at all? Authority: may it act without a human? Both are enforced by the platform itself — an AI cannot talk its way past either.

The three authority levels

• Autonomous — the employee acts on its own, including outside-impact actions like sending messages.
• Approval — internal work happens automatically, but anything with outside impact (SMS, email, sending documents, payment links, dispatching, changing job status) is held for you.
• Suggest — the employee may only read and recommend; it can never cause outside impact.

The approvals queue

Held actions appear under Pending approvals on the AI Workforce page with the exact action, its arguments, the employee's reason, and an expiry (7 days). Approving executes exactly that stored action, exactly once — even if two admins click at the same moment. Rejecting is final. Expired approvals can no longer execute.

Our recommendation

Start new employees on Approval, watch a week of proposals, then grant Autonomous per employee once you trust its judgment. The Command Center defaults to Approval for outside-impact actions.

Still stuck?

Support and the contact form reach a human.