Most AI projects go wrong in the same quiet way: nobody decided in advance what the AI is allowed to do. So the AI drafts the strategy, invents the experience, picks the price, and hits publish — while the human becomes a rubber stamp for work they never actually reviewed. This prompt builds the missing document before work begins: a role-division table for any project, generated from a short description of what you are trying to do. The principle it encodes comes from a real AI-agent project and is simple: the human keeps decisions, money, identity, and final approval; the AI does research, options, drafting, and bookkeeping. The prompt interviews you about your project, then produces a complete table — task by task, who owns it, what the AI may do without asking, what it must stop and escalate, and what it must never touch. It also generates the escalation rules and the "hard boundaries" list (payments, identity documents, publishing, external messages) in plain language you can paste into any agent instruction or team doc. You get the prompt, the reasoning behind each boundary, and a filled example from a content-monetization project. For anyone delegating real work to AI agents, assistants, or tools — and wanting the delegation to stay safe, reviewable, and genuinely theirs.

What you get

- The complete role-division table template prompt

- A task-by-task format for ownership, permissions, escalation, and never-do boundaries

- Default principles that keep decisions, money, identity, and final approval with the human

- A filled example from a content-monetization project