Practical Task
Turn a complicated project idea into the first five concrete steps
Stop asking AI for a giant master plan. Give it the outcome and constraints, then force the project down to actions you can start this week.
What this helps you do
Anyone with a project that feels too large or fuzzy to begin.
What you need
- The project idea
- Why you want it
- Constraints
- What you already have or know
Do it step by step
- 1
Describe the outcome you want in normal language and why it matters.
- 2
Add constraints such as budget, deadline, people involved, tools, or things you refuse to compromise on.
- 3
Ask AI to separate unknowns that truly block progress from details that can wait.
- 4
Request only the first five concrete actions, each small enough to complete or clearly start.
- 5
After doing the first few steps, return with what you learned and re-plan from reality rather than expanding the original hypothetical master plan.
Use this in your AI assistant
Starter instructions
Copy the starter below, then paste it into ChatGPT, Copilot, Gemini, Claude, or whichever capable AI assistant you use. If this Task calls for photos or files, attach them in that AI assistant too — not on this page.
Help me start this project without giving me a 50-step master plan. Outcome: [OUTCOME]. Why it matters: [WHY]. Constraints: [CONSTRAINTS]. What I already have/know: [CONTEXT]. First identify only the unknowns that genuinely block progress, then give me the first 5 concrete actions in order. Each step should produce something visible or answer one important question.
Helpful checkWhat a good result looks like
Five actions that make the project less hypothetical and give you new information for the next planning round.
Tips and watch-outs
- If step one is 'research,' ask the AI to define exactly what question the research must answer.
- Do not let planning become a substitute for doing one reversible first step.
OptionalGo deeper when you want to
Separate assumptions from facts before the plan grows
Large projects become brittle when the first brainstorm quietly turns assumptions into architecture. Ask the assistant to keep a short assumption log: what you know, what you think is probably true, and what must be tested before later decisions depend on it.
Not every unknown deserves research. Prioritize assumptions that could materially change the next few actions.
Favor steps that buy information cheaply
The strongest early step is often reversible and produces evidence: talk to three users, price one dependency, build a throwaway prototype, test one import, measure one process, or try one small version manually.
Ask the AI to explain what each step is supposed to teach you and what result would change the next move.
Information-value test
For each of the first five steps, add: what evidence this step should produce, the assumption it tests, and what result would make us change direction. Prefer reversible steps that teach us something before expensive commitments.
Re-plan from reality at a deliberate checkpoint
After two or three steps, compare the original assumptions with what actually happened. Delete steps that no longer make sense and add only the next few actions you can now justify.
A good AI planning loop gets shorter and more evidence-based over time. It should not keep expanding the original hypothetical master plan just because more detail is available.