Units
1. What is Artificial Intelligence, Machine Learning and LLM? 2. How Does Generative AI Work? 3. What Can AI Do and What It Can't Do? 4. Fundamentals of Effective Prompt Writing 5. Hallucination and the Limits of Credibility 6. Prejudice and Justice 7. Data Privacy and KVKK: Where Can I Write Company Data? 8. Everyday AI Use Scenarios in the Workplace 9. End-to-End AI Workflow: Draft, Validate, Decide 10. Copyright, Ownership and Transparency 11. Responsible Use and Human Control 12. Personal AI Roadmap and Continuous Learning
Unit 4 / 12

Fundamentals of Effective Prompt Writing

Gains:

  • Can list the basic components that a good prompt should have.
  • Can improve output quality by giving role, context, task and format
  • Can turn a weak prompt into a strong prompt

The same AI tool produces completely different results in the hands of two different people. The difference is not in the tool, but in the instructions. The instruction you give to the AI ​​is called a prompt. Writing a good prompt is a learnable skill, not magic; In fact, this is the most practical benefit of this module. In this unit, you will learn the basic components that turn a weak instruction into a strong instruction, with copyable patterns. Good news: a few simple habits will instantly double your output quality.

Five Components of a Good Prompt

A powerful prompt usually carries these five elements. You don't have to use them all every time, but the more you give, the more accurate the result will be.

  1. Role: Tell the model which expert you want it to act as. (“You are an experienced customer service professional.”)
  2. Context: Give the situation, background, target audience. ("An angry customer has been waiting for a response for 3 days.")
  3. Task: Be clear about what you want. (“Draft a response that includes an apology and offers a solution.”)
  4. Format: Specify the format of the output. (“5 sentences maximum, no bullet points, formal but warm tone.”)
  5. Constraints: Set limits. ("Don't promise a discount, don't give a date, don't add a signature.")
Tip: When writing a prompt, ask yourself: "If I gave this task to a new intern, what would I need to tell him or her to do the job correctly?" All that information should go into the prompt. The model, just like that intern, makes the most general prediction with incomplete information.

Weak Prompt / Strong Prompt

Weak prompt: Write an apology email to the customer.

Result: A general, cliché text as it is not clear who, for what, and in what tone.

Powerful prompt:Role: You are an experienced customer service professional.Context: A customer complained because his shipment was 4 days late. A loyal, long-term customer.Task: Draft a response that offers a brief apology and a clear solution.Format: 6 sentences or less, friendly but professional tone.Restriction: Do not promise a refund; Just say "I forwarded the situation to the resolution team".

Result: A draft that is contextualised and very close to use.

Step by Step: Developing a Prompt

Writing a prompt is not a one-time process, but an iterative (improved step by step) process. If the output doesn't suit your liking, rewrite; This is not a failure, it is the method itself.

  1. Write the first draft, print it out.
  2. Figure out what you don't like (is it too long? is the tone wrong? is it missing?).
  3. Add a single instruction that will fix this: “make it shorter,” “switch to a more formal tone,” “add an example.”
  4. Repeat until you like it; Then save this prompt.
Caution: Do not ask the model for too many things at once. Massive requests like “write a marketing plan, 10 slogans, a budget sheet, and an email” produce messy output. Divide the big job into parts; Ask for one clear thing at every step.

Three Mini Cases

Case 1 — 3 words to 3 minutes. A sales rep typed "write offer email" and received a useless text, spent 20 minutes correcting it by hand. The next day he used the role + context + format template; The output was almost ready, sent it in 3 minutes. Same tool, different instruction.

Case 2 — The power of specifying formatting. An HR professional wanted to summarize 60 candidate feedback. Got a mixed paragraph on the first try. When I said "Pour the result into a 3-column table: theme, how many people, sample comments", a usable, presentable table appeared. Difference: saying the desired format.

Case 3 — Effect of giving example. An employee couldn't get the email tone he wanted. Finally, he pasted in an old email he liked and said, "Write it in this tone." The model captured the style. Giving an example (few-shot) is often more powerful than describing.

Copiable Templates

Role: [what expert to act like]Context: [situation, background, target audience]Task: [exactly what you want]Format: [length, tone, structure]Constraint: [what not to do]

Improve text below. Give me:1) Edited version2) Give me 3 important changes you made and why. Tone: [e.g. official]. Length: [e.g. half as much].Text: [paste here]

I give an example that I like; Generate a new text in the same style and tone.Example: [paste the text you like]New topic: [new topic]

Produce me 3 different versions of this task:1) short and formal2) friendly and medium length3) detailed and persuasiveTask: [task]. Then say in one sentence which one you recommend and why.

Comparison Chart

Weak prompt feature

Strong prompt response

"Write something" (indefinite task)

A clear, single task

Audience not specified

Target audience defined

Tone unspoken

Tone and style specified

Length/format free

The desired format is clear

Many requests at once

Steps broken into pieces

Common Mistakes

Common mistakes

  • Not giving context. The model does not know you or your situation; If you don't say who you are and why you want what you want, you will get a general answer.
  • Not specifying the format. Instructions such as "Let it be a table", "Let it be 5 items", "Do not exceed 3 sentences" make the output usable.
  • Giving up at the first exit. If the first result is not liked, the prompt is improved; The process is iterative.
  • Asking for everything with a single prompt. Divide the big task; Ask for one clear thing at every step.
  • Forgetting to give an example. It is often more effective to show an example rather than describe a style.

In summary

  • A good prompt usually contains role, context, task, format and constraint components.
  • “What would I tell a new intern?” The question indicates what information you will provide.
  • Writing a prompt is an iterative process; If the first output is not liked, improve the instruction.
  • Break large tasks into parts; Don't ask for too much at once.
  • Demonstrating a style by example is often more powerful than describing it.

Application Task

Think back to a task you recently asked the AI for and got poor results. Rewrite it with the "Role / Context / Task / Format / Constraint" template and put the two outputs side by side. Save the new prompt that works in a notes file for future use.

Checklist

  • [ ] I can name five components of a good prompt.
  • [ ] I can improve the output quality by adding roles, context and format.
  • [ ] I can turn a weak prompt into a strong prompt.
  • [ ] I know that writing a prompt is an iterative process.
  • [ ] I save useful prompts and use them again.
Previous Next