Prompt Engineering for Beginners: How to Get Better Results From AI

Introduction

AI Is Only as Good as the Instructions You Give It. Prompt engineering isn't about finding a magical phrase that makes AI perfect.

It's about giving AI enough context, constraints and direction to understand what you're actually trying to accomplish.

The 5 Parts of a Good Prompt

Teach a simple framework:

R — Role Tell the AI what perspective or expertise it should use. "Act as a senior frontend developer..."

T — Task Clearly explain what you want.

"Create a responsive landing page..."

C — Context

Give it the information it needs.

"The website is for a freelance marketplace targeting African freelancers..."

C — Constraints Tell it what it should and shouldn't do. "Use vanilla JavaScript. Don't use React or external libraries."

O — Output Specify what you want the final answer to look like. "Return the HTML, CSS and JavaScript in separate code blocks and explain how to connect them."

You could call this the RTCCO framework or simplify it into a memorable formula: Role + Task + Context + Constraints + Output

Turn a Bad Prompt Into a Good Prompt

This would be one of the strongest sections. Example 1 — Coding

Bad: Fix my JavaScript.

Better: I'm building a todo list using vanilla JavaScript. When I click the delete button, it doesn't remove the correct todo. Here is my HTML and JavaScript: [code]. Identify the exact problem, explain why it's happening, then provide the smallest possible fix. Don't rewrite the entire application.

Then explain why this is better.

Example 2 — Writing

Bad: Write a LinkedIn post about JavaScript.

Better: Write a LinkedIn post about completing my JavaScript learning journey. I'm a frontend developer looking for remote opportunities. Keep it professional but personal, around 150 words, and avoid generic motivational language.

Example 3 — Research

Bad: Tell me about React.

Better: I'm a frontend developer who already knows HTML, CSS and vanilla JavaScript. Explain React specifically in terms of what problem it solves compared with vanilla JavaScript. Use simple examples and assume I understand DOM manipulation but haven't used React before.

This shows readers that the more relevant context you give the AI, the more relevant the response becomes.

Don't Ask AI to Do Everything at Once

This is an important practical lesson.

Instead of: Build me a complete e-commerce website with authentication, payments, admin dashboard, database and deployment.

Break the project into stages.

Step 1 Help me plan the architecture.

Step 2 Create the homepage.

Step 3 Create the product page.

Step 4 Add the cart functionality.

Step 5 Review the code for bugs.

This gives you much more control.

Give AI Examples of What You Want

One of the easiest ways to improve an AI response is to show it an example.

Instead of: Make the text professional.

Say: Write it in this style: [example]. Keep the same level of professionalism and sentence length, but don't copy the wording.

This is particularly useful for:

• Social media posts
• Blog posts
• UI copy
• Emails
• Marketing
• Code formatting

Tell AI What NOT to Do

People often focus only on what they want. But constraints can be just as important.

For example: Build this using vanilla JavaScript. is useful.

But: Build this using vanilla JavaScript. Do not use React, jQuery, Bootstrap or other frameworks. is much clearer.

You can also specify:

• Don't rewrite working code
• Don't add unnecessary dependencies
• Don't invent information
• Don't change the existing design
• Don't shorten the explanation
• Don't use technical jargon

Use AI as a Collaborator, Not Just a Generator

This would be a great section for Wurk users.

Instead of: Give me a business idea. Use AI throughout the process: Give me 10 ideas.

Then: Rank these ideas based on startup cost and difficulty.

Then: Challenge the strongest idea. What could make it fail?

Then: Give me a simple MVP.

Then: Create a 30-day execution plan.

You're no longer just asking AI for an answer. You're using it to think through a problem step by step.

Ask AI to Critique Its Own Work

After getting an answer, don't immediately accept it.

Ask: Review your previous answer. Identify three weaknesses or assumptions you made and explain how you would improve them.

For code: Review this code as a senior developer. Look specifically for bugs, security issues, unnecessary complexity and poor practices.

This can catch problems that aren't obvious at first glance.

Give AI Your Actual Material

This is probably one of the biggest differences between useful AI and generic AI.

Don't say: Help me fix my website.

Give it: Your code The error message Your expected result What actually happened Relevant screenshots

For example:

Expected: Clicking Delete should remove only the selected task. Actual: Clicking Delete always removes the first task. Code: [paste code]

Now the AI has something concrete to reason about.

A Prompt Template You Can Reuse

End the article with something readers can copy.

ROLE: Act as a [type of expert].

TASK: I need you to [specific task].

CONTEXT: Here is the relevant information: [provide information]

CONSTRAINTS:

• Use [technology/style]
• Don't [specific restriction]
• Keep it [length/complexity/etc.]

OUTPUT: Give me the result in [specific format].

QUALITY CHECK: Before giving me the final answer, check for [errors/bugs/inconsistencies].

Then give a complete example using the template.

Conclusion

The biggest lesson I'd want readers to leave with is:

Don't try to find the perfect prompt. Learn how to communicate clearly with AI.

A good prompt gives AI the information it needs to understand:

Who it should act as → what you want → why you need it → what limitations exist → what the final result should look like.

And when the first answer isn't good enough, don't immediately start over.

Refine the prompt.

Tell the AI what was wrong, provide more context, add constraints, show an example, and try again.

That's the real skill behind prompt engineering.