Building Blocks for Better Prompts: A Modular Prompt Engineering Framework

Anyone who’s explored the ups and downs of learning to work with generative artificial intelligence (GenAI)—particularly large language models (LLMs) like ChatGPT, Gemini, or Claude—knows how intimidating prompt engineering can seem. To make it more accessible, let’s adopt a creative framework that turns complexity into play: a building-blocks metaphor.

Why building blocks? Because, just like those colorful bricks can build anything from castles to spaceships, prompt components can be assembled to produce anything from bullet-point summaries to strategic business proposals.

In this guide, you’ll find a practical framework for writing prompts (with examples) using three levels of “bricks” or prompt elements:

• Basic bricks — your standard toolkit for almost every prompt
• Intermediate bricks — to enhance your “construction”
• Advanced bricks — to get the most out of your interaction with AI

Rather than just listing components, this guide walks you through a prompt-building journey that follows the natural flow of creating powerful interactions. Throughout the steps, capitalized words (e.g., TASK, ROLE, CONTEXT) represent the individual “bricks” or prompt elements in this framework.

Before you build: How to use this prompt engineering guide

This isn’t an exhaustive encyclopedia of prompt techniques. Think of it as a modular starter kit—a curated set of the most useful bricks for common use cases.

A quick word of caution: while the building-blocks metaphor is helpful, it shouldn’t be taken too literally. Prompt components aren’t always as neat and compatible as real bricks. The interaction between elements can be unpredictable, and results may vary even with the same “build.”

Unlike a real set of toy bricks—where following the same instructions produces the same result every time—GenAI is probabilistic. Even identical prompts can yield different outputs because randomness is built into the system. Embrace that uncertainty and treat prompting as a creative exploration rather than a strict formula.

Ready to get started? Let’s begin your prompt-building journey.

Step 1: Define what you want the AI to do

State the TASK (Basic)

This is the heart of your prompt. You must clearly tell the AI what you want it to do:

• "Summarize this text."
• "Generate meeting minutes from the attached transcript."
• "Advise me on how to make effective presentations."

Modern models are often smart enough to give you a decent response with just this core request. But if the output isn’t quite right—too generic, off-target, or simply underwhelming—it’s time to reach for your bricks and start building a better prompt.

When to add more bricks

LLMs generate responses based on patterns in the data they were trained on. If your request is too vague, the answer will default to the most common, safe, and middle-of-the-road content.

Want something specific—like insights from the perspective of a software developer in the telecom industry? You need to guide the model toward that particular "neighborhood" of its training data. That’s when you start adding more bricks to steer the model toward your goal.

Assign a ROLE (Basic)

The role tells the AI what kind of thinker, expert, or collaborator to emulate: Editor, Innovator, Expert, Mentor, Coach, Prompt Generator, Sparring Partner, Project Manager… you name it.

• "Act as a management coach facilitating a brainstorming session."
• "Be a strategy advisor analyzing market trends."
• "Speak like a skeptical engineer testing an idea."

The more clearly defined the role, the better the model can tune its output.

Step 2: Give the AI what it needs to accomplish the task

Provide CONTEXT (Basic)

The richer the context, the more targeted the output. Providing context helps the AI lay out the most relevant trajectory through its vast knowledge. You can:

• Explain your current situation, problem, task or objective
• Specify your industry, project background, product features, etc.
• Add relevant details

Think of AI as a new colleague who wants to help but knows absolutely nothing about your world. What would you need to tell that person?

Describe your INPUTS (Basic)

If you’re feeding the AI some content—whether text, a transcript, or a table—briefly explain what it is and what to do with it. Mention:

• File name or structure: "This is a transcript from a project kick-off call. ", or "The attached file ’Transcript.docx‘ contains the transcript from…"
• Key fields or content type: "The attached file contains answers to a questionnaire, separated by a header ’Response X‘ and containing three fields: name, org unit, response text."
• What you expect the AI to extract or transform: "Focus on user feedback about feature X."

Give EXAMPLES (Intermediate)

Just as the picture on the box shows the completed model, examples guide AI toward your desired output. Provide samples of:

• Preferred formatting or structure
• Writing style or tone
• Solution approaches
• Template to follow

This helps AI "see” the finished model you're aiming for.

Assign a PERSONALITY (Intermediate)

LLMs are surprisingly good at adopting personality traits. Add color by asking the model to be "imaginative," "detail-oriented," "skeptical," or even to channel specific MBTI types (e.g., "Think like an INTJ strategist" or "Respond like an ESTP doer").

Use this when attitude or cognitive style matter for the task.

Establish CONSTRAINTS and CONDITIONS (Intermediate)

Narrow down what you don’t want or must adhere to:

• "Exclude technical jargon."
• "Only use facts from the provided document."
• "Focus only on European regulations."

This helps focus the model’s attention and keeps it from wandering off.

Ask for an INTERVIEW (Advanced)

When you’re not sure how to frame your request, ask the AI to interview you. Let it gather context systematically and adapt follow-up questions based on your responses. To keep the exchange manageable, you can tell the AI to ask you one question at a time—and let it know to stop once the key points are covered.

Use case examples:

• Planning an article when you're unsure where to start
• Clarifying complex input
• Collaboratively shaping a plan

Think of it as having an AI assistant who helps you gather all the bricks needed for your build.

Step 3: Explain how you want the AI to respond

Specify the OUTPUT FORMAT (Basic)

Be precise about the structure you need:

• "Respond with a concise paragraph."
• "Consolidate findings into a numbered list."
• "Give me a table with 3 columns: 1) A, 2) B, 3) C."

Set the LENGTH and DETAIL (Basic)

Help control verbosity and precision:

• "In five bullet points"
• "No more than 350 words"
• "Just one-sentence summary”

Clear boundaries help AI deliver precisely what you need.

Adapt to your AUDIENCE (Intermediate)

Tailor the output to the person or group it’s meant for:

• "Make it understandable to non-technical personnel."
• "Use business language for executives."
• "Explain as if to a skeptical engineer."

Select the STYLE and TONE (Intermediate)

Match the style to your purpose. Should it be professional, non-technical, brief, elaborate, executive-ready?

Tone tweaks how the message feels emotionally: Formal, playful, persuasive, narrative, academic.

Choose the LANGUAGE (Intermediate)

AI is your multilingual companion! You can:

• Input in your preferred language, get output in another
• Ask for a translation
• Request multiple language versions simultaneously
• Specify dialect preferences (US/UK English, etc.)

Request MULTIPLE OPTIONS (Advanced)

Like spreading out different builds, asking AI for multiple options helps you:

• Compare different approaches
• Mix and match elements you like
• Discover new perspectives

Request variations by specifying:

• Different angles (“Show me 3 approaches: technical, business, and creative”)
• Various formats (“Present this as a list, a story, and a dialogue”)
• Multiple levels (“Give me a basic, intermediate, and advanced version”)

Step 4: Ask the AI to reflect and refine

Ask the AI to EXPLAIN its reasoning (Intermediate)

Ask AI to clarify its work:

• "Walk me through your reasoning."
• "What factors influenced this recommendation?"
• "How did you arrive at this conclusion?"

This helps you understand the AI's thought process and validates its suggestions.

Request SELF-REFLECTION (Advanced)

Ask the AI to assess or critique its own output:

• "Rate the quality of your answer"
• "What assumptions did you make?"
• "Critique your suggestions for clarity, feasibility, and potential impact"

This approach is great for refining drafts, exploring biases, evaluating robustness or stress-testing results or assumptions.

Invite IMPROVEMENT (Advanced)

Let the AI act as its own editor or coach:

• "Suggest how to improve this email."
• "What’s missing in this proposal?"

You can even combine this with a ROLE or PERSONALITY brick: "How would an engineer / lawyer / data scientist rewrite this?"

Prompting Examples in Action: From Simple to Advanced

To see how prompt components stack together, here’s one scenario—asking AI to help improve project risk management— shown at increasing levels of complexity, where each added “brick” improves clarity, specificity, and usefulness.

Task only

“Suggest ways to improve project risk management.”

Task + Role + Format

“As a project risk management consultant, list a few practical ways to strengthen risk management in projects. Present results as a prioritized list in tabular format.”

Task + Role + Context + Input + Personality + Format + Audience

“You’re a project management coach. Suggest improvements to risk management based on the attached risk management section of the recent project audit report of my process automation project.

Be pragmatic and structured. Format your recommendations in a four-column table: Priority, Improvement Action, Rationale and Implementation Effort. Target the audience of non-risk management professionals.”

Task + Role + Context + Input + Personality + Constraints + Format + Detail + Audience + Style + Multiple Options + Self-reflection

“You are a project management consultant tasked with strengthening risk management practices across the portfolio of process automation projects.

Use the attached summary of risk management sections of recent audit reports (RM_Summary_2025.docx), structured by categories: risk management governance, risk response practices, to evaluate current approaches. Adopt a pragmatic and creative perspective. Avoid suggesting new governance layers or measures requiring additional personnel.

Provide three key improvements in a table: Idea, Description, Impact, and Implementation Difficulty. Limit descriptions to 20 words.

Write in a professional, decision-ready style for executive audience. Offer two sets of ideas: 1) traditional and 2) AI-enabled.

After that, reflect on the clarity of each idea, its practical relevance, and how easily it could be implemented in complex process automation projects.”

Finishing touches: Advice for better prompts

Think first. Prompt second.

Prompting works best when you know what you want. GenAI can’t replace thinking—it enhances it. So, take a moment to clarify the problem you’re solving or the outcome you need. Otherwise, you might get a beautifully worded answer…to the wrong question.

Keep it simple

Yes, prompts are powerful, but that doesn’t mean they need to be complex. Use only the bricks you need for the task. Sometimes, just a core task is enough.

Match your objective

Think about what you’re trying to achieve, then pick bricks accordingly. Content creation? Focus on output format, style, and audience. Problem-solving? Load up on context, constraints, and reflective bricks. The right combination makes all the difference.

Learn by doing

Prompting is a skill—and a creative one! As you're starting out, follow the steps and recipes suggested here. But as you gain experience, don’t be afraid to mix it up. Try swapping the order, combining unusual bricks, or testing new combinations. Your masterpiece might start where the instructions end.

Remember, effective prompting is an iterative process. As you become more experienced, you'll develop an intuition for which techniques work best in different situations.

NOTE: This material was created through a human-led process, with GenAI supporting as a co-intelligence partner— helping generate ideas, refine and proof content, and offer critical perspective.

Tags: Artificial Intelligence | Generative AI | Digital Transformation | Optimization | Future of Work

Related Insights

You May Also Like