Human-in-the-Loop

Human-in-the-loop is a design pattern where AI prepares the work but a person approves it before anything happens. The system drafts, suggests, and queues; you decide what actually goes out.

It matters most where mistakes are public — a sent email can't be unsent. Gatherly applies it everywhere it acts on your behalf: reply drafts wait for your sign-off, and on WhatsApp nothing leaves until you reply "send draft 1". An AI chief of staff should earn trust per message, not assume it up front.