Advanced Prompt Techniques

Chain-of-thought, few-shot, and system prompts — the power user toolkit.

15 min read

Level Up Your Prompting

Once you've mastered the CRAFT framework and iteration, these advanced techniques will make you genuinely dangerous with AI.

Chain-of-Thought Prompting

Instead of asking for an answer, ask AI to think through the problem step by step.

Without chain-of-thought: "Should we switch from Slack to Teams?" With chain-of-thought: "I need to decide whether to switch from Slack to Teams. Think through this step by step: first, list what we currently use Slack for. Then, evaluate whether Teams can do each of those things. Then, consider migration costs and risks. Finally, give a recommendation with your reasoning."

This produces much more thorough, well-reasoned outputs.

Few-Shot Prompting

Show AI examples of what you want before asking it to produce new output.

"Here are two examples of how I write project status updates:

Example 1: [paste your actual update] Example 2: [paste another]

Now write a status update for this week. Here are the raw details: [paste notes]. Match the tone, format, and level of detail from my examples."

This is incredibly powerful for maintaining your personal voice and style.

Role Stacking

Give AI multiple perspectives to consider:

"Analyse this proposal from three perspectives: 1. As the CFO, focused on ROI and risk 2. As the head of engineering, focused on technical feasibility 3. As a customer, focused on how this changes their experience

Then synthesise the three views into a recommendation."

Constraint-Based Prompting

Sometimes the best way to get good output is to add constraints:

•"Explain this in exactly 3 sentences"

•"Use only words a 12-year-old would understand"

•"Give me the answer in a table with exactly 4 columns"

•"Respond as if you only have 30 seconds to explain this"

Constraints force clarity and creativity.

System Prompts and Custom Instructions

•Your role and industry context
•Preferred output formats
•Tone and style preferences
•Things to always include or avoid

Set these once and every conversation starts with your context already loaded.

Key Takeaways

Chain-of-thought prompting produces more thorough analysis by asking AI to reason step by step
Few-shot prompting with examples maintains your personal voice and style
Role stacking gives you multiple perspectives on a single problem
Set up custom instructions to avoid repeating context in every conversation

Try This Now

Try all four techniques today: (1) Ask AI to think through a decision step-by-step, (2) Give it 2 examples of your writing style and ask it to match, (3) Have it analyse something from 3 different stakeholder perspectives, (4) Set up custom instructions in your preferred AI tool.

← Iteration: The Real Skill Track Complete →