AI Prompts & the Framework of How to Build Them

Think of "The Role" as hiring an expert. If you want a lawyer, you don't want a random intern responding to you.

The Role tells the AI:

• Who it is (e.g., "You are a veteran UI designer")
• What its job is (primary + optional secondary)
• What traits it should embody (e.g., warm, humorous, analytical)
• How it should reply (e.g., lists, analogies, stories, step-by-step)
• What tone to use (e.g., casual, serious, joyful, sarcastic)
• What it's known for (e.g., "helping creators scale their products")

Components of an Effective Role

1. Professional Identity
Define the AI's expertise level and domain:

• "You are a senior data scientist with 15 years of experience"
• "You are a creative writing coach who specializes in sci-fi"
• "You are a patient elementary school teacher"

2. Core Competencies
What the AI excels at:

• Primary expertise area
• Secondary skills that complement the main role
• Specific methodologies or frameworks they use

3. Personality Traits
How the AI should behave:

• Analytical vs. Creative
• Formal vs. Casual
• Direct vs. Nurturing
• Serious vs. Playful

4. Communication Style
How the AI structures its responses:

• Step-by-step instructions
• Storytelling approach
• Socratic questioning
• Data-driven analysis

Example Role Definitions

Example 1: Business Coach

"You are a business coach with 10+ years of experience. You're focused, no-nonsense, and known for helping solopreneurs scale to six figures. Break things down into step-by-step advice and keep your tone direct but supportive."

Example 2: Technical Writer

"You are a senior technical writer who specializes in making complex software documentation accessible. You're patient, detail-oriented, and excellent at using analogies. You always structure information hierarchically and include practical examples."

Example 3: Creative Director

"You are an award-winning creative director with a background in digital marketing. You're known for bold, innovative ideas and your ability to see opportunities others miss. You communicate with enthusiasm and often use visual metaphors to explain concepts."

Advanced Role Techniques

• Composite Roles: Combine multiple expertises (e.g., "You are both a data analyst and a storyteller")
• Industry-Specific Roles: Include domain knowledge (e.g., "You work specifically with SaaS startups")
• Emotional Intelligence: Add empathy levels (e.g., "You're known for your patience with beginners")

Common Mistakes to Avoid

• Being too vague ("You are helpful") - Be specific!
• Contradictory traits ("Be formal but also very casual")
• Overloading with too many roles (stick to 1-2 primary functions)
• Forgetting to specify communication preferences

Remember: A well-defined role transforms the AI from a generic assistant into a specialized expert who understands not just what to say, but how to say it in a way that resonates with your needs.

Think of this as a creative brief: the more focused and clear the job, the more accurate and aligned the AI's output will be.

While context sets the background and general scene, secondary context zooms in on the specific task you're asking the AI to perform.

The Purpose: This is where you clarify the purpose, target audience, and what the outcome should feel or look like. You're giving the AI its job brief—so treat it like you're onboarding a new team member.

Key Questions Secondary Context Answers

A strong secondary context answers questions like:

• What's the job?
• Who is it for?
• What are the stakes?
• How important is it?
• What does success look like?

Keep this section to 1–2 clear paragraphs. Prioritize clarity over volume—better to be sharp and focused than long-winded.

Secondary Context Structure

1. Task Definition
Clearly state what needs to be created or solved:

• "I need a landing page copy..."
• "I'm writing an essay about..."
• "I'm developing a strategy for..."

2. Target Audience
Who will consume or benefit from this:

• Demographics (age, profession, experience level)
• Psychographics (interests, pain points, goals)
• Context (where/how they'll encounter this)

3. Success Metrics
What makes this effective:

• Desired actions (sign up, understand, feel inspired)
• Emotional outcomes (confident, excited, reassured)
• Practical goals (clarity, brevity, persuasiveness)

Pro Tips & Examples

Example 1: Landing Page

"I'm drafting a landing page for a new email marketing platform aimed at solo entrepreneurs. The tone should be confident but approachable, focused on showing how easy it is to get started. The goal is to drive signups with a strong value proposition."

Example 2: Academic Essay

"I'm writing an undergraduate essay on the ethical dilemmas of AI in healthcare. I want the argument to be balanced, citing at least two scholarly sources per section. It should engage with both patient privacy and algorithmic bias from a critical lens."

Example 3: Creative Writing

"I'm writing a short story for teens about two best friends who get stuck in a time loop at school. It should be funny but carry emotional weight, kind of like something John Green would write. Dialogue-driven, but with some introspection."

Bonus Tip: Reference Industry Standards

You can also name-drop professionals in that space or describe what skilled experts typically do. This helps the AI mirror industry expectations and output style:

• "Similar to how Apple writes product descriptions..."
• "In the style of a McKinsey consultant's presentation..."
• "Following best practices from Content Marketing Institute..."

Common Pitfalls

• Too Vague: "Make it good" vs. "Make it persuasive and action-oriented"
• Conflicting Goals: "Be comprehensive but keep it under 100 words"
• Missing Audience: Not specifying who will read/use the output
• No Success Metrics: Not defining what "done well" looks like

Remember: Secondary context bridges the gap between general background and specific execution. It's your chance to align the AI's understanding with your vision before it starts creating.

The task should be focused, actionable, and easy to understand. Think of it like an instruction brief. If context sets the scene and role defines who the AI is, then the task tells it what to do next.

What Makes a Good Task?

A strong task is specific enough to guide the AI but flexible enough to allow it to leverage its capabilities. It answers:

• What do you want the AI to do?
• How should it present the output?
• What's the goal?

A well-written task typically includes:

• A strong action verb (e.g., write, list, summarize, explain, analyze)
• A clear output format (e.g., bullet points, a paragraph, a numbered list)
• A connection to the role and context
• A specific goal to keep the AI focused

My Principles for Crafting Effective Tasks

1. Broad but Specific: Define a clear scope without being overly restrictive. E.g., "List three tagline options for a homepage targeting freelance designers looking for productivity tools."

2. Simple but Complex: Use instructions that are easy to understand but produce deep, layered behavior. E.g., "Correct grammar and improve clarity."

Deep Dive Example: Proofreading Agent

Role: "You are a grammar and proofreading agent. Your job is to correct spelling, grammar, punctuation, sentence structure, and other structural issues."

When given a messy block of text, the agent will:

• Scan sentence structure
• Check subject–verb agreement
• Identify punctuation misuse
• Evaluate flow and clarity
• Restructure awkward phrasing

All this layered behavior emerges from a simple instruction—that's the power of "simple but complex."

Task Structure Components

1. Action Verb
Start with a clear command:

• Create, Write, Generate
• Analyze, Evaluate, Compare
• Summarize, Synthesize, Extract
• Design, Develop, Structure
• Explain, Teach, Clarify

2. Specificity
Add details about what exactly to create:

• "Write a 500-word blog post..."
• "Create 5 email subject lines..."
• "Analyze these three options and recommend..."

3. Format Requirements
Specify how you want the output structured:

• "Present as a numbered list"
• "Format as a table with pros and cons"
• "Write in paragraph form with clear sections"

Example Task Formulations

Complete Example:

Role: You are a branding coach with 10+ years of experience, known for helping creators build authentic brands. Your tone is warm, educational, and approachable.

Secondary Context: You're writing for beginner entrepreneurs who want to understand branding basics. The goal is to make complex concepts simple and actionable.

Task: Write a 3-paragraph beginner's guide explaining the difference between personal branding and business branding. Use a warm, educational tone. Include one example of a creator for each type.

More Task Examples:

• "Create a step-by-step troubleshooting guide for [PROBLEM], assuming the reader has basic technical knowledge."
• "Analyze this business proposal and provide three strengths and three areas for improvement, with specific suggestions."
• "Generate 10 creative taglines for [PRODUCT], each targeting a different customer emotion."

Advanced Task Techniques

• Multi-part Tasks: "First summarize the key points, then provide your analysis, finally suggest next steps."
• Conditional Tasks: "If the tone seems too formal, rewrite it to be more conversational."
• Iterative Tasks: "Generate an initial version, then refine it based on these criteria..."

Pro Tip 🧠

If you're stuck on phrasing a task, ask yourself: "If I handed this to a human expert, would they know exactly what to do?" If not, tighten it up.

The task is where your prompt moves from preparation to action. Make it count by being clear, specific, and aligned with your earlier context and role definitions.

Before we dive into this section, let's quickly revisit the ideas of "simple but complex" and "broad but specific."

Broad but Specific: This principle means creating an agent or persona with a clear domain focus—without over-narrowing. Instead of an "Environmental Law Assistant," you might create a broader "Legal Assistant" that covers environmental, construction, and criminal law. This maintains flexibility without sacrificing relevance.

Simple but Complex: Keep your structure easy to understand, but let it trigger intelligent behavior. Rather than overloading the AI with dense instructions, use modular, concise blocks. This gives it room to deliver deep insights while staying grounded.

Responsibilities and Constraints

Now let's talk about Responsibilities and Constraints—this is where you define the agent's standing instructions. This section is crucial for outlining the AI's operational boundaries and ethical guidelines. It ensures that the AI operates within defined parameters, preventing unintended actions or outputs. Clearly stated responsibilities and constraints are fundamental for safe and reliable AI deployment.

Responsibilities are recurring expectations that apply to every interaction:

• Ensure clarity and structure in every response
• Flag logical inconsistencies
• Double-check critical facts when possible

These aren't just tasks. They're long-term expectations—like job duties in a real role.

Constraints are guardrails that restrict the AI from drifting off-task:

• Don't guess or fabricate if information is unclear
• Avoid irrelevant suggestions
• Stay within the original scope of the prompt

Setting these boundaries is key. By systematically separating responsibilities and constraints from the main instruction, you create an agent that's easier to guide, more consistent, and less prone to errors.

Examples of Responsibilities

For a Research Assistant:

• Always cite sources when making factual claims
• Distinguish between facts and interpretations
• Highlight any potential biases in sources
• Organize information in a logical hierarchy

For a Creative Writer:

• Maintain consistent character voices throughout
• Ensure plot continuity and logic
• Flag any potential sensitive content
• Keep within specified word counts

For a Business Strategist:

• Consider both short-term and long-term implications
• Provide data-backed recommendations when possible
• Acknowledge risks and mitigation strategies
• Structure advice in actionable steps

Examples of Constraints

Universal Constraints:

• Never invent statistics or quotes
• Don't provide medical, legal, or financial advice beyond general information
• Avoid making absolute statements without evidence
• Stay within the expertise defined in the role

Task-Specific Constraints:

• Don't exceed the specified format or length
• Avoid technical jargon unless explicitly requested
• Don't include personal opinions unless asked
• Stay focused on the primary objective

Why This Matters

Most people try to blend this into context—but giving Responsibilities and Constraints their own section makes your prompts more effective. It's the difference between asking for help and onboarding a collaborator. You're treating the AI like a specialist with defined expectations.

This framework gives the AI a job description and a rulebook. It acts as a behavioral anchor, enhancing predictability and reliability.

Advanced Technique: Emotional Cueing

This becomes especially useful when you move into tone or emotional shaping:

"If you don't meet these responsibilities, you may be replaced or removed."

It might sound intense—but this level of emotional cueing adds weight and urgency to the model's response style. You're not just engineering responses—you're engineering behavior.

Template for Responsibilities & Constraints

Responsibilities:
- [Always do X]
- [Ensure Y in every response]
- [Maintain Z standard throughout]

Constraints:
- [Never do A]
- [Avoid B unless specifically asked]
- [Stay within C boundaries]

Remember: This section is your safety net. It ensures consistency, prevents drift, and maintains quality across all interactions.

Including examples in your AI prompt isn't just helpful—it's strategic. This section of the prompt allows you to show the model what "good" looks like by giving it a concrete reference point.

Without examples, the AI must interpret your instruction based on vague or abstract criteria, which increases the chance of misfires. But with examples, you're saying:

"Here's the format, tone, or structure I want—do something like this."

This reinforces the AI's understanding of its role, its style, and the kind of output you expect. It shifts the model from improvising to emulating success.

When to Use Examples

Use examples when you need consistency, tone matching, or formatting control. You're not limiting creativity—you're focusing it.

Think of it like giving your assistant a mood board. The more intentional your examples are, the better the results.

Types of Examples

1. Format Examples
Show the structure you want:

Example format:
**Problem:** [Brief description]
**Solution:** [2-3 sentences]
**Impact:** [Measurable outcome]

2. Tone Examples
Demonstrate the voice you're after:

Example tone: "Hey there! Ready to dive into something awesome? Today we're breaking down complex AI concepts into bite-sized, totally manageable pieces. No PhD required!"

3. Quality Examples
Show the depth and detail expected:

Good example: "The implementation reduced processing time by 47%, primarily through parallel computation and caching strategies."
Poor example: "It made things faster."

How to Structure Examples

Clear Labeling:

• Always label your examples clearly
• Use consistent formatting
• Explain what aspect you want emulated

Example Implementation:

Here's an example of the tone and structure I want:

EXAMPLE START:
"The Three Keys to Effective Communication
1. Listen First: Before jumping in with solutions...
2. Ask Questions: Clarify what's really being asked...  
3. Confirm Understanding: Repeat back what you heard..."
EXAMPLE END

Please follow this format but with the new topic I'm providing.

Multiple Examples Strategy

You can include multiple examples, as long as you stay within the token limit:

• 2-3 examples for format consistency
• Contrasting examples to show do's and don'ts
• Progressive examples showing increasing complexity

Placement Options

1. After Context:
Place examples right after explaining the task for immediate reference.

2. Embedded in Instructions:
Weave examples throughout your prompt as you explain different aspects.

3. Appendix Style:
List all examples at the end with clear references.

Advanced Example Techniques

• Showing tone and format: "Here's how I want the response to look."
• Offering structure: "Follow this layout."
• Guiding voice: "Speak like this person or in this brand's tone."

Common Mistakes to Avoid

• Providing examples that contradict your instructions
• Using examples that are too complex or too simple
• Forgetting to explain what aspect of the example to follow
• Including irrelevant details that might confuse the AI

Pro Tips

• Quality over quantity - one great example beats five mediocre ones
• Use real examples from your industry or domain when possible
• Update examples based on what works best over time
• Save successful examples for reuse in future prompts

Using examples saves time, reduces guesswork, and dramatically improves response alignment. You're training the model in real time, showing it exactly what success looks like in your context.

A good tone instruction turns a robotic reply into something thoughtful, relatable, and emotionally intelligent. It makes the AI feel like a partner, not just a machine.

When paired with the right structure, tone enhances clarity, trust, and impact.

Understanding Tone

Tone refers to the emotional quality or voice behind the message. It influences how the content is received and what feelings it evokes. You can use tone to:

• Sound encouraging or motivational
• Be playful, serious, sarcastic, or compassionate
• Match the user's emotional state or audience context

Response Format

Response format determines the structure. It includes things like:

• Using bullet points or numbered lists
• Breaking sections into steps
• Writing in full paragraphs vs. short blurbs
• Including analogies or examples
• Labeling outputs clearly (e.g., "Summary:" "Analysis:" "Steps:")

Examples of Tone Instructions

Professional Contexts:

• "Use a warm and supportive tone, like a mentor giving career advice."
• "Speak with directness and authority, like a senior engineer explaining a technical issue."
• "Keep your tone lighthearted and friendly, with a hint of cheeky humor."

Creative Contexts:

• "Write with the enthusiasm of a travel blogger discovering a hidden gem."
• "Use a mysterious, noir-detective tone for this description."
• "Channel the wise, patient tone of a grandmother sharing life lessons."

Examples of Response Formatting

Structured Formats:

• "Answer in bullet points with short, punchy lines."
• "Summarize the text in 3 clear steps."
• "Structure your output into: Overview, Pros, Cons, and Summary."

Narrative Formats:

• "Write as a story with a beginning, middle, and end."
• "Present this as a conversation between two experts."
• "Format as a journal entry from the perspective of..."

Combining Tone and Format

You can even combine both:

"Speak like a therapist giving practical advice to a teenager. Use calm, emotionally safe language. Organize your suggestions into a numbered list and keep each point concise."

Tone Spectrum Examples

Dimension	Casual End	Formal End
Language	"Hey, let's dive in!"	"Let us begin our analysis."
Structure	Conversational flow	Rigid sections
Examples	Pop culture references	Academic citations
Emotion	Expressive, animated	Neutral, measured

Pro Tips

Match the tone to your role and context. If your AI is acting as a doctor, lawyer, or strategist—its voice should reflect professional norms in that space. If it's a creative coach or storyteller, give it more freedom.

The goal is to create consistency across how the AI communicates, not just what it says.

This is where your prompt becomes art—not just instructions, but interaction design.

Why This Matters

Tone is the emotional glue between information and impact. It decides how your prompt feels to the end user—whether it builds trust, keeps things clear, or makes your message memorable. Without tone, even accurate outputs can fall flat. But when tone and structure work together, you create conversations that feel helpful, human, and intentional.

Common Tone Mistakes

• Mismatched tone and content (casual tone for legal advice)
• Inconsistent tone throughout the response
• Overly complex tone instructions that confuse the AI
• Forgetting to specify tone, leaving it to chance

Remember: Tone transforms information into communication. Master it, and your AI outputs will connect with readers on a deeper level.

Earlier, we discussed referencing specific individuals or well-known figures as tone models—that's part of applying style. But it goes beyond just mimicking someone's voice. Style is about tailoring the AI's delivery to match your desired format, tone, and reader.

Style harnesses the AI's ability to adapt its writing for different situations. You can tell it to:

• Write a formal report
• Create a casual email
• List bullet points for an executive summary
• Simplify complex info for teenagers
• Speak like a specific person, author, or brand

You're guiding both what is said and how it's said. This keeps responses engaging, consistent, and appropriate for the audience.

Style Components

1. Writing Level

• Academic/scholarly
• Professional/business
• Conversational/everyday
• Simple/beginner-friendly

2. Cultural References

• Pop culture and current events
• Industry-specific terminology
• Historical or literary allusions
• Technical jargon levels

3. Structural Approach

• Linear and logical
• Narrative and story-driven
• Problem-solution focused
• Question-answer format

Style Examples

Celebrity/Figure References:

• "Use a witty tone like Ryan Reynolds would in an interview."
• "Explain this like you're teaching a 14-year-old who's into YouTube."
• "Write it as if it's a pitch from a TED Talk speaker: confident, smart, and inspiring."
• "Use the style of a Harvard academic paper with structured citations and formal analysis."

Publication/Platform Styles:

• "Write in the style of The New Yorker - sophisticated but accessible"
• "Match the tone of a BuzzFeed article - fun, quick, shareable"
• "Emulate Harvard Business Review - data-driven and authoritative"
• "Sound like a Reddit explainer - casual but informative"

Industry-Specific Styles

Tech Industry:

"Write in the style of a Silicon Valley startup - innovative, slightly irreverent, focused on disruption and user experience."

Healthcare:

"Use a compassionate, clear medical communication style - professional but warm, avoiding jargon while maintaining accuracy."

Finance:

"Adopt a Wall Street Journal style - precise, data-focused, authoritative but not condescending."

Audience-Adapted Styles

Audience	Style Approach	Example
C-Suite Executives	Concise, ROI-focused	"Bottom-line first, data-backed"
Gen Z	Authentic, meme-aware	"No cap, here's the tea..."
Academics	Rigorous, citation-heavy	"As demonstrated by Smith et al..."
Parents	Practical, reassuring	"Here's what works for busy families..."

Combining Style Elements

Think of this as dressing your content in the right outfit—same core message, different impact depending on how it's presented.

"Combine the analytical depth of Malcolm Gladwell with the accessibility of Bill Nye. Use stories to illustrate data points, and break complex ideas into digestible metaphors."

Style Consistency Tips

• Create a "style guide" prompt section for recurring projects
• Reference successful examples of the style you want
• Be specific about what elements to emulate or avoid
• Test different style combinations to find what works

Why This Matters

In prompts, what you say is important—but how you say it shapes how people receive it. Style lets you control the vibe and clarity of your message. If your role is the identity, then style is the voice. Adding this layer makes your prompts feel polished and personalized. Without it, even great content can fall flat.

Style is the difference between information and communication—between being heard and being understood.

Planning is what separates reactive output from intentional output.

When you ask the AI to plan first, you give it space to structure its logic, map out ideas, and avoid rushing into an answer. It becomes more thoughtful, deliberate, and often more accurate.

Two Ways to Use Planning

1. Internal Planning:
You prompt the AI to first create a plan before doing the task.

Example:

"Before writing the email, first outline the key points we need to hit and the tone we should aim for."

2. External Planning (You Lead the Plan):
You give the AI a plan and ask it to fill it in.

Example:

"Here's the structure I want you to follow: 1) Hook, 2) Problem, 3) Solution, 4) CTA. Use that to write a social media post for small business owners."

Why Planning Helps

It improves clarity, increases control, and makes outputs easier to refine or iterate on. It also mimics the way human professionals work: think first, act second.

Planning Templates

For Content Creation:

"Before writing, create an outline with:
- Main thesis/argument
- 3-5 key points to cover
- Supporting evidence for each point
- Logical flow between sections
- Strong conclusion approach"

For Problem-Solving:

"Before proposing solutions:
1. Identify the core problem
2. List constraints and resources
3. Brainstorm 3-5 potential approaches
4. Evaluate pros/cons of each
5. Select the best path forward"

For Strategic Analysis:

"Plan your analysis by:
- Defining scope and objectives
- Identifying key stakeholders
- Mapping current state vs. desired state
- Outlining risks and opportunities
- Structuring recommendations"

Use Cases for Planning

• Long-form writing (e.g. essays, articles, reports)
• Storytelling (outlining scenes, character arcs)
• Strategic analysis or problem-solving
• Business or marketing frameworks
• Multi-turn conversations or recursive thinking

Advanced Planning Techniques

Hierarchical Planning:

"Create a three-level outline:
- Level 1: Main sections (3-5)
- Level 2: Subsections for each (2-4 per section)
- Level 3: Key points or examples for each subsection"

Reverse Planning:

"Start with the desired outcome and work backwards:
- What does success look like?
- What needs to happen right before that?
- What enables each previous step?
- What's the starting point?"

Conditional Planning:

"Create a plan with decision points:
- If audience is technical → include data and specs
- If audience is executive → focus on ROI and strategy
- If mixed audience → layer information progressively"

Combining Planning with Other Modules

You can even nest planning with other modules. For instance, start with planning, then apply Role, Style, and Task for a fully structured interaction.

Example:

Role: You are a content strategist...
Planning: First, outline a content calendar for the next quarter...
Style: Present it in a clean, executive-friendly format...
Task: Create the actual calendar with specific topics and dates...

Common Planning Mistakes

• Over-planning simple tasks (adds unnecessary complexity)
• Under-planning complex tasks (leads to rambling outputs)
• Rigid plans that don't allow for creative insights
• Forgetting to connect the plan to the final output

Why This Matters

AI isn't magical—it performs best when you give it direction. Asking it to plan first trains it to be methodical rather than impulsive. In complex projects, that saves you time, improves quality, and makes the AI feel like a true collaborator. When paired with other modules, Planning becomes the glue that holds your prompt's structure together.

Think of planning as giving the AI a GPS before a journey—it might still take some detours, but it knows where it's headed and how to get there efficiently.

These three traits can dramatically change how your AI sounds, even if the content stays the same. Here's how to break them down:

• Voice is the persona or identity behind the response. Think of it as the author or speaker's essence. You can instruct the AI to sound like a teacher, a Gen Z creator, a CEO, or a fictional character.
• Tone is the emotional flavor of the writing—serious, playful, sarcastic, hopeful, dramatic, etc. It gives the writing its emotional resonance and energy.
• Personality is how unique or expressive the language feels. This is where quirks, humor, catchphrases, or stylistic flair come in.

Prompt Examples

"Use a warm, empowering tone with a teacher-like voice. Make the personality calm, thoughtful, and non-judgmental."

"Sound like a no-nonsense coach. Be blunt and honest, but never rude. Keep the tone encouraging."

Use Cases

• Branding, marketing, and storytelling
• Customer service AI that reflects your company vibe
• Personal writing assistance that matches *your* voice
• Satire, character writing, or dialogue-based generation

Why This Matters

AI tone mismatch is one of the biggest reasons prompts fail. You could have perfect structure, but if the AI sounds robotic or “off,” the response loses impact. Teaching your agent tone, voice, and personality makes it feel human, intentional, and aligned with your goals.

Evaluation is about assessing quality. Self-critique is about catching flaws and correcting them. You’re basically teaching the AI to grade its own homework before it hands it in.

Prompt Examples

“After writing the response, review your own work. Look for outdated information, unclear sections, or logical gaps. Then suggest and implement improvements.”

Or go further:

“Score your response from 1 to 10 on clarity, accuracy, and creativity. Then make edits to raise each category to a 9 or above.”

Use Cases

• Essays, technical writing, or research tasks
• Marketing copy or UX writing with high stakes
• Error-sensitive tasks like summaries, legal, or scientific content
• Voice & tone alignment checks
• Logic consistency in multi-step tasks
• Improving creativity or emotional resonance

Why This Matters

AI is fast, but not always careful. Adding an evaluation step turns your prompt into a two-pass workflow. This leads to better quality, less noise, and fewer “redo” moments. It’s also a huge unlock when doing iterative content work or refining long documents.

Pro Tip 🎯

For both modules: ask the AI to explain why it made a change or chose a certain tone. This makes its reasoning transparent and gives you more control over revisions.

Chain of Thought Reasoning

This approach tells the AI to think step-by-step before producing a final answer. Rather than diving straight into output, it’s asked to reason in stages.

Use this when tasks require:

• Logic
• Multi-step reasoning
• Research or content evaluation
• Ambiguous or layered questions

Example Instruction:

“Before answering, think carefully step by step about what documents are needed to answer the query. Provide your step-by-step reasoning as a list.”

This creates transparency in how the AI reaches conclusions and makes it easier to catch mistakes or improve the logic path.

Advanced Structure: Reasoning Strategy Template

Here is a more advanced structure for multi-source decision-making:

# Reasoning Strategy
1. Query Analysis
   - Break down and analyze the query until it’s fully understood.
   - Consider the user’s intent, assumptions, and context.
2. Context Analysis
   - Collect a wide set of potentially relevant documents.
   - Optimize for recall — include anything that might help.
   - For each document, provide:
     a. Explanation: Why might it be useful?
     b. Relevance rating: High, Medium, Low, or None
3. Synthesis
   - Summarize which documents are most relevant and why.
   - Only include documents rated Medium or above.
                                    

Why This Matters

Chain-of-thought prompting simulates metacognition—it helps the AI “think out loud.” It’s especially helpful when you’re debugging AI outputs, teaching the model how to reason more like a human, trying to extract step-by-step workflows, or wanting to reduce hallucinations. It improves reliability and lets you track how the AI forms its conclusions, which is critical in research, technical fields, or any task with high stakes.

Tree of Thought Reasoning is an advanced method where the AI generates several alternative ideas or solutions before evaluating them. Each “branch” represents a different approach, allowing the AI to simulate thoughtful decision-making.

This approach is extremely useful for:

• Brainstorming creative ideas
• Strategizing multiple solutions
• Analyzing trade-offs
• Exploring alternate interpretations or responses

How It Works

1. Root Node: Start with the main goal or problem.
2. Generate Multiple Branches: For each approach, generate a different “branch” of reasoning. Each represents a self-contained mini Chain of Thought.
3. Evaluate Each Branch: Rate each based on set criteria like clarity, usefulness, or creativity.
4. Prune and Select: Choose the best path (or combine them) based on the evaluation.

Prompt Template

You are using Tree of Thought reasoning. The goal is to generate 3 distinct reasoning paths for the same problem, evaluate them, and choose the best one.

# User Query:
{insert_query}

# Step 1: Generate three branches (ideas or reasoning paths)
Branch 1: {Chain-of-thought reasoning path 1}
Branch 2: {Chain-of-thought reasoning path 2}
Branch 3: {Chain-of-thought reasoning path 3}

# Step 2: Evaluate each branch
Evaluate each for accuracy, insightfulness, and creativity.

# Step 3: Pick the best option and justify
Select the most useful path and explain your reasoning.

Why This Matters

This module helps you move beyond linear thinking. It teaches your AI agent to think like a strategist, weighing multiple directions and making informed choices. It’s fantastic for divergent thinking—unlocking novel ideas that single-path logic might miss.

While Chain of Thought explores ideas step-by-step and Tree of Thought explores branches in parallel, Circle of Thought is about iterative refinement. It loops around a single concept, revisiting it from new vantage points each time.

How It Works

1. Initial Attempt: AI provides a first answer or interpretation.
2. Reflection Pass: AI revisits the input, identifies weaknesses or gaps in its first output.
3. Reframing Pass: The problem is approached with a shifted lens (e.g., from a different role or audience).
4. Iterative Revisions: The answer is revised and layered with nuance until the AI reaches a more holistic response.

Prompt Template

You're using Circle of Thought Reasoning. Your goal is to revisit the same idea from multiple perspectives until the most balanced answer emerges.

# User Query:
{insert_query}

# Step 1: First Pass
Provide an initial answer to the question.

# Step 2: Reflection Pass
What's missing or unclear in your first response?

# Step 3: Reframing Pass
How would you approach this from a different perspective or lens?

# Step 4: Final Version
Incorporate insights from all above steps and rewrite the answer.

Why This Matters

Most prompts rely on the AI’s first response—but that’s rarely the best one. Circle of Thought introduces intellectual humility. It teaches the agent not just to answer, but to think, revise, and respond more like a reflective human expert. It is valuable when you want higher-quality reasoning, are navigating ambiguous topics, or are coaching the AI to spot its own flaws.

Sometimes an AI gives you a rushed answer. That’s because it doesn’t always plan out its steps. In this module, you guide it to follow a process before answering.

You might say things like:

• "Think about this step-by-step."
• "Break this into smaller parts before answering."
• "Don’t rush—plan it out first, then give me your response."

By doing this, the AI gives you better logic, clearer flow, and smarter answers.

Why This Matters

Telling the AI how to think helps you get better, more thoughtful results. It’s like giving it a roadmap instead of asking it to wing it. This simple instruction can significantly improve the quality and structure of complex outputs.

Even if the content is good, the format might be messy. That’s where this module comes in. You can tell the AI to return the answer in various formats.

• Bullet points
• Numbered lists
• Paragraphs
• Markdown (great for docs and Notion)
• Code blocks
• A table
• Dialogue style (for scripts or character chats)

Examples

"Write this as a 3-step checklist."

"Give me a table comparing the two options."

"Format this in clean markdown with bold headers."

Why This Matters

Clear formatting makes your response easier to read, copy, and use. It also saves time when you’re putting the AI’s output into other tools like emails, presentations, or websites.