Layered Prompting Framework

What You Get

A step-by-step method to break complex tasks into phases instead of asking AI for the final answer in one shot. You’ll get better output, more control, and less wasted tokens.

Why This Works

Most people ask Claude: “Write me a lead generation blog post about productivity tools.” Then they get something generic back because Claude has no anchors.

Layered prompting gives Claude anchors at every stage. You’re not asking for a finished product — you’re collaborating on one, the way you’d brief a writer or designer on a team.

The five layers:

1. Clarify the objective — What are we actually solving for?
2. Break down the strategy — What’s the architecture?
3. Execute in phases — Build piece by piece.
4. Expand with detail — Fill in the depth.
5. Refine and finalize — Polish and test.

The Prompt

I'm working on a complex project and I want to build it in layers instead of asking for everything at once.

Here's my task: [YOUR TASK HERE]

Before you generate anything, walk me through the layered approach:

1. CLARIFY: Ask me 3-4 clarifying questions about the objective, audience, constraints, and success metrics.
2. STRATEGY: Once I answer, propose a 3-5 step breakdown of how you'd approach this. Include what we'll tackle first, what depends on what, and what final refinements will look like.
3. EXECUTION: Don't generate the full output yet. Ask me: "Ready to move to Phase 1?" Then build the first phase, pause for my feedback, and repeat.
4. EXPANSION: Once we have the core, we'll add depth — detail, examples, specificity, voice.
5. REFINEMENT: Final pass for clarity, impact, and fit.

Let's start with step 1. Ask me your clarifying questions.

How to Use It

Step 1: Start the conversation. Paste the prompt above with your task filled in. Answer Claude’s clarifying questions. Be specific about who this is for and what success looks like.

Step 2: Get the strategy breakdown. Claude will propose a 3-5 step approach. Review it. If something’s out of order or missing, say so. (“Actually, we need to nail the headline before the outline” or “This is good but add a step for audience research.”)

Step 3: Execute phase by phase. When you say “ready for phase 1,” Claude builds the first piece. Review. Give feedback. Then move to phase 2. This is where iteration happens — you’re steering, not just receiving.

Step 4: Add depth. Once the structure’s solid, you’ll add detail, examples, specificity. This is faster because the hard thinking is done.

Step 5: Refine. Final pass for voice, flow, impact. This usually takes minutes, not hours.

Example Output

A Real Layered Project: Lead Generation Blog Post

Your task: “Write a blog post that generates leads for my AI coaching service. Target audience: founders and solopreneurs who use AI but feel scattered.”

Claude’s clarifying questions:

1. What’s your biggest competitive angle? (What makes your coaching different from generic AI advice?)
2. What action do you want readers to take? (Sign up for a free guide? Schedule a call? Join a workshop?)
3. What’s a common mistake or pain point these founders experience with AI?
4. Do you have any stats, case studies, or specific wins you want to reference?

Your answers: “My angle is systematic — I teach founders how to architect their AI workflow, not just prompt better. I want them to grab a free ‘Founder’s AI Operating Manual’ PDF. The pain point is that they’re using AI for everything but have no system, so they’re either overcomplicating it or missing opportunities. I have 3-4 client wins I can anonymize.”

Claude’s strategy breakdown:

1. Hook + Problem — Open with the pain point. Land it hard.
2. The Opportunity — Shift to what’s possible with a system.
3. The Method — Introduce your 5-step framework (high level).
4. Social Proof — One or two anonymized wins.
5. Call to Action — Lead magnet + benefit.

“Ready for Phase 1?”

Phase 1 output:

# Stop Using AI Like You're Hiring 50 Freelancers

You're smart. You're using Claude, ChatGPT, all of it. But you're using it like each tool is a separate contractor — asking for a blog post here, analysis there, a script over here. No handoff. No consistency. No system.

The result? You get output. But you don't get leverage. You're not architecting your work — you're just outsourcing pieces of it.

Here's what I'm seeing with founders right now: They're spending 5 hours a week on AI tasks that should take 1. They're regenerating the same prompts across 3 different projects. They're losing context between tools. They have 47 documents and no way to find what they built last month.

This isn't an AI problem. It's an architecture problem.

You read Phase 1. You note: “This is good, but it’s a bit dark. Can you lift the tone slightly in paragraph 2? I want them to feel possible, not stuck.”

Claude adjusts. You continue through each phase.

Tips

One: Ask clarifying questions before generating. This is the difference between good output and great output. The questions force you to think clearly, and Claude gets the information it needs.

Two: Use the “power prompting technique” at any layer. If Claude is giving you one approach, ask for three with pros/cons before you pick one. This takes 30 seconds and often surfaces something better.

Example: “Before you write Phase 2, give me three different structural approaches to this section with pros and cons for each. Then tell me which one you’d recommend and why.”

Three: Pause between layers. Don’t rush from strategy to execution. Let the structure sink in. Sometimes the best feedback comes after you’ve sat with the plan for 10 minutes.

Want more methods? Browse the full library at hazelq.com/methods. Built with Claude. Every prompt in this library has been tested in the latest Claude model.