← Back to Learn AI

Advanced Techniques

You've got the basics. Now let's get dangerous. 🔥

Jump to:

Chain of Thought•Role Playing•Multi-Perspective•Structured Outputs•Multiple Tools•Context
🔄

Chain of Thought Prompting

For complex problems, ask AI to show its reasoning step by step. This dramatically improves accuracy on difficult tasks and helps you understand the logic behind the answer.

When to use:

Strategy decisions, technical architecture, debugging complex issues, financial analysis, anything where you need to understand the "why" behind the answer.

How to use it:

Add phrases like "Think through this step by step," "Walk me through your reasoning," or "Break this down into steps" to your prompts.

Example Prompt:

I need to decide whether to build this feature in-house or buy a third-party solution. Walk me through your reasoning step by step. Consider: cost, timeline, maintenance, customization needs, and team capacity. Show your work.

🎭

Role Playing

Ask AI to take on a specific role or perspective. This changes how it approaches problems and gives you feedback from angles you might not have considered.

When to use:

Getting feedback from different angles, stress-testing ideas, preparing for difficult conversations, understanding stakeholder perspectives, interview prep.

Effective roles:

  • "Act as a skeptical investor" - tough questions about viability
  • "Respond as an experienced product designer" - UX/design feedback
  • "You're a customer who tried this product" - user perspective
  • "Be my debate opponent" - find holes in arguments
  • "Act as a technical architect" - infrastructure decisions

Example Prompt:

Act as a skeptical board member. I'm presenting this strategy: [paste your plan]. Ask me the hard questions an experienced executive would ask. Push back on weak spots. Make me defend every assumption.

🔍

Multi-Perspective Analysis

Get AI to analyze the same problem from multiple viewpoints in one response. This surfaces blind spots and helps you make more balanced decisions.

When to use:

Major decisions, contentious issues, cross-functional projects, product strategy, anything where different stakeholders have different priorities.

How to structure it:

Ask for analysis from 3-5 specific perspectives. Be explicit about what each perspective should prioritize.

Example Prompt:

Analyze this product decision from three perspectives:
1. Engineering: technical complexity, maintainability, tech debt
2. Business: revenue impact, market position, competitive advantage
3. User: UX, friction points, value delivered
For each perspective, tell me the biggest concerns and biggest opportunities. Then recommend a path forward that balances all three.

📊

Structured Outputs

Request specific formats (JSON, tables, bullet lists) so outputs can be easily used in other tools or workflows. Essential for automation and integration.

When to use:

Building tools, data analysis, API integrations, spreadsheet work, creating databases, any time you need to move AI output into another system.

Common formats:

  • JSON: For APIs and programming
  • Tables: For spreadsheets and reports
  • Markdown: For documentation
  • CSV: For data processing

Example Prompt:

Extract key information from this sales call transcript and return as JSON with this structure:
{
  "customer_name": "",
  "pain_points": [],
  "budget": "",
  "timeline": "",
  "next_steps": []
}

🤝

Using Multiple AI Tools Together

Here's a technique most people miss: use different AI tools like coworkers with different skills. Each tool has strengths — leverage them together.

The mental model:

Think of each AI as a specialist on your team. Perplexity is your researcher. Claude is your architect. ChatGPT is your writer. You're the manager deciding who does what.

Powerful combos:

  • Perplexity → Claude: Research with Perplexity (it cites sources), then bring findings to Claude for synthesis and strategy
  • ChatGPT → Claude: Brainstorm ideas fast with ChatGPT, then have Claude pressure-test and build the best one
  • Claude → ChatGPT: Have Claude write a detailed prompt, then paste it into ChatGPT for execution
  • AI critiquing AI: Paste Tool A's output into Tool B and ask "What's wrong with this? What's missing? How would you improve it?"

Example Workflow:

1. Ask Perplexity: "What are the latest trends in [industry]? Cite sources."
2. Copy that research into Claude: "Based on this research, write a strategy memo for my team."
3. Ask ChatGPT: "Make this memo more engaging and add a punchy executive summary."

🔥 Power move:

Ask one AI to write a prompt for another AI. "Write me a detailed prompt I can give to ChatGPT to generate 10 LinkedIn post ideas about [topic]." The prompt it writes will be better than what you'd write yourself.

📁

Context Management

AI doesn’t remember you from past conversations. Each new chat starts from zero unless memory is turned on. Even then, it may be missing context. The fastest way to level up your results? Give AI the context it needs upfront. Save key details you want AI to remember and then provide that context each session.

The problem:

Without context, AI gives generic answers. With context, it gives answers tailored to your situation, your goals, and your constraints.

What to include in your context:

  • Who you are: Role, industry, experience level - leave out personal or sensitive information
  • What you're working on: Current projects, goals, deadlines - do not share confidential information
  • How you like to work: Communication style, format preferences
  • What you've tried: Past approaches, what worked/didn't
🔥 Power move:

Create a context document you can paste at the start of important chats. Save it somewhere easy to grab. I've created a template you can steal.

📄 Get the Context Template

🧪 Combining Techniques

These techniques are even more powerful when used together. Here's how to layer them:

Chain of Thought + Role Play

"Act as a CFO. Walk through your reasoning on this budget allocation step by step."

Multi-Perspective + Structured Output

"Analyze from 3 perspectives and return as a comparison table with pros/cons for each."

All Four Together

"Act as three stakeholders. Each should think through this step by step and provide feedback as structured bullet points."

⚠️ When NOT to Use These

Advanced techniques add complexity and use more tokens. Use them strategically:

  • Simple tasks: Don't overcomplicate. "Write a professional email" doesn't need chain of thought.
  • Creative work: Too much structure can kill creativity. Let AI flow naturally for brainstorming.
  • Quick answers: If you just need a fast fact, keep it simple.
  • When you're experimenting: Start simple, then add techniques if basic prompts aren't working.
Rule of thumb:

If the task is important, complex, or involves tradeoffs between multiple priorities, use advanced techniques. For everything else, keep it simple.

Want to see these in action? 🎯

Check out the projects I've built using these techniques, or take a brain break.

📊 See My Data🎮 Fun with AI
← Back to Learn AI

A note from Jamie: This is my personal take on AI, based on my experience using it. This is general information, not professional advice. Use this as a guide, not a rulebook.

Built by Jamie at buildwithjamie.com