What Is Prompt Engineering? An Essential Skill for the AI Era
Prompt engineering is the skill of instructing AI to get usable results. Learn the 5-part prompt structure, advanced techniques, before-and-after examples and cautions.
Two people on the same team use the same AI tool. One gets an email draft that is almost ready to send. The other complains the AI gives vague answers that are useless. The difference is not the tool, it is how they give instructions. This skill is called prompt engineering, and it is not just for tech people. Anyone who writes emails, prepares reports or plans work benefits.
What prompt engineering is
A prompt is the text you type to instruct the AI. Prompt engineering is designing that text so the AI understands the context, the task and the output format you want. Think of briefing a new team member. If you only say make some slides, you will get work that needs heavy editing. If you say who will see it, what it is for and how many pages, the result comes much closer to what you need. AI works the same way.
Before and after
Prompt that is too short: Write an email.
Result: a generic email with no idea who it is for, which needs to be almost fully rewritten.
A better prompt: Write a follow-up email to a client who asked for a quotation 5 days ago. Polite tone but nudging them to decide, no more than 5 lines, ending with an invitation to a short phone call.
Result: a usable draft that only needs a name and a few details adjusted.
A 5-part prompt structure
1. Role: set the role
For example You are a marketer who understands the Thai market. This helps the AI choose the right perspective and language level.
2. Context: give background
For example An organic skincare shop whose main customers are working women, selling mainly through TikTok Shop. The clearer the context, the more relevant the answer.
3. Task: state the job clearly
Use clear verbs such as write, summarise, compare or analyse, and say what the output will be used for.
4. Format: define the shape
For example 5 bullet points, a table, 3 options or no more than 200 words.
5. Constraints: state the limits
For example use simple language, no jargon, do not cite numbers without a source. Constraints prevent outputs you do not want.
3 techniques used often in real work
Few-shot: show examples first
Paste 2 to 3 examples you like, then ask for 5 more in the same style. This works well for captions, product descriptions and customer replies. Read more in few-shot prompting.
Ask for step-by-step reasoning
For tasks that need reasoning, such as comparing options or planning, ask the AI to lay out its reasoning before concluding, so you can check where the answer came from. Some newer models already reason step by step on their own, but asking to see the reasoning still helps you verify it.
Refine in rounds
Do not expect a great result on the first try. Just say good, but cut it in half, or too formal, make it friendlier. This back-and-forth often works better than writing one very long prompt.
Cautions
- AI can be confidently wrong: always check numbers, names, laws and facts against real sources.
- Confidential data: do not paste customer data, personal data or confidential company documents unless you are sure of your company's policy.
- You still own the work: AI helps draft, but you are the one who submits it.
AI answers only as well as you can explain the task.
Where to start
- Pick one repetitive weekly task, such as meeting summaries or email drafts.
- Write a prompt using the 5-part structure and try it for real.
- Refine it 2 to 3 times until you are happy with the result.
- Save prompts that work into a library. See a prompt library for teams.
- Keep learning with the courses below.
Courses on Plants.work Learn to go further: Learn prompt engineering to communicate with AI effectively, Getting to know prompt engineering and 5 business prompts to sell more with ChatGPT and Gemini. Or practise asking career questions with the AI Career Assistant and notice how much better the answers get when your question is clearer.