Imagine working with someone who can answer questions, but can only communicate through notes slipped under a door. You carry the information in. You carry the answer out. Every part of turning it into something useful is still yours.
That’s the starting image for this episode. The six levels describe what changes as you give AI more context, tools, continuity, and responsibility.
It’s my working model of delegation, not an industry standard or a ranking of people’s ability. A single tool can support several of these patterns, and different parts of your work can belong at different levels.
The map
| Level | What changes | Your job |
|---|---|---|
| 1. Conversation | You bring the material in and take the answer out. | Supply context and check the answer. |
| 2. Tools | It can work with an approved file or service. | Choose access and review its actions. |
| 3. Continuity | Instructions or memory carry useful context forward. | Correct what is saved and keep it current. |
| 4. Specialists | Different roles handle research, drafting, or review. | Route the work and manage handoffs. |
| 5. Coordination | A lead role moves work between specialists. | Set the brief and inspect the process. |
| 6. Bounded autonomy | Work proceeds between your reviews. | Set budgets, stopping rules, and evidence requirements. |
The building metaphor helps me see the trade. More capability can mean more useful work gets done. It also means there is more happening between my instructions and the result I see.
These aren’t rigid technical boundaries. For example, a chat tool may already have memory, and several roles can coordinate without using the same connection standard. The question is what your particular setup actually does.
The awkward middle
At the time of recording, I described myself as working around level four and moving toward five.
That middle stage can create more work. You have a researcher, a writer, and a reviewer. You are carrying material between all three. A new role only earns its place if the quality or saved effort justifies that coordination.
Before adding another specialist, ask where the current arrangement is failing. Sometimes a better brief or a saved example is enough.
Choose one small rep
Use the delegation worksheet to map one task you already do. Name what AI can access, what it may change, and how you will judge the result.
Then try one small extension if it solves an actual problem. Let it organize a copy of a file. Save an instruction you keep repeating. Give a draft to a separate reviewer and compare the criticism with the original evidence.
The goal is to hand over work you can still assess. The Homebrew Club business experiment shows what can happen when a reassuring report gets ahead of the evidence behind it.