Learning Prompt Engineering Through Real Work Examples in Singapore
Why Practice With Real Examples Matters
Prompt engineering is the skill of writing clear instructions for language models to get useful output. Many tutorials explain theory, but the fastest way to learn is to work through actual tasks. By applying prompt engineering to real work examples, you build intuition for what works and what doesn't.
This approach is especially valuable for professionals in Singapore who want to use AI tools in their daily work. Whether you are drafting reports, generating code, or summarising documents, practicing with examples from your own field helps you understand how to structure prompts for consistent results.
Starting With a Simple Task
Pick a task you already do manually. For example, writing a short product description or a summary of a meeting. Write out the task as a prompt without any special structure. Just describe what you want in natural language. Run it and note the output.
What to watch: does the model give you exactly what you asked for? Often the first attempt is too vague. The model might add extra details, miss key points, or produce text that feels generic. That is normal. The goal is to see where the prompt falls short.
Add Structure to Your Prompt
Now rewrite the same task with a clear structure. Start with a role. For example, “You are a professional copywriter.” Then state the task. “Write a product description for a wireless keyboard. Keep it under 100 words. Use a friendly tone.” Add an output format. “Return only the description, no introductory text.”
What to watch: does the output match the format and tone you specified? Role prompting often improves relevance. If the model still drifts, check whether your instruction is ambiguous. Replace vague words like “friendly” with concrete examples: “Use short sentences and casual words like ‘great’ and ‘easy’.”
United Ceres College is also on Facebook.
Using Few Shot Examples
For tasks where you need a specific style or pattern, provide a few shot examples. Suppose you want the model to write customer replies in a standard format. Give it two or three examples of the input and the desired output. Then present the new input.
What to watch: the examples must be representative. If your examples are all short and the real input is long, the model might truncate its reply. Also, check that the pattern you want is clear from the examples alone. If the model tries to copy the content rather than the structure, adjust the examples or add an instruction like “Follow the structure above.”
Managing the Context Window
Every model has a maximum context window, the amount of text it can process at once. When working with long documents or multi-step tasks, you need to fit your prompt, examples, and any reference material within that window.
What to watch: if the model seems to forget earlier instructions, you might be exceeding the context window. Break the task into smaller steps. Alternatively, put the most critical instructions at the very end of the prompt, because some models pay more attention to the last part of the input.
Defining Evaluation Criteria
Before you iterate, decide how you will judge success. Write down three to five criteria. For example, accuracy, tone match, length, and whether the output includes all required sections. Run the prompt and score the output against each criterion.
What to watch: be specific about what counts as a pass. “Accurate” is too broad. Instead, say “The description must not contain any technical errors about the product.” This makes iteration objective. When you adjust the prompt, you can compare scores to see if the change helped.
Iterating Systematically
Change only one element of the prompt at a time. That might be the role, the instruction clarity, the number of examples, or the output format. Run the prompt again and note the score. If the score improves, keep the change. If not, revert and try something else.
What to watch: keep a log of each version and its score. It is easy to forget what you tried. A simple text file with the prompt and the output is enough. Over several rounds, you will see which types of changes have the biggest effect.
This process of structured iteration is exactly what you learn in a good AI prompting classes environment. The focus is on practical application rather than abstract theory. By working through your own examples, you internalise the principles faster.
Common Pitfalls and How to Avoid Them
One common mistake is overloading the prompt. When you add too many instructions, the model may miss some. Keep each prompt focused on one main task. If you need multiple outputs, write separate prompts.

Another pitfall is assuming the model knows your context. For example, if you ask for a summary of a meeting but do not include the meeting notes, the model will guess. Always provide the relevant text in the prompt. Use quotation marks or a separator to distinguish between instructions and content.
A third issue is inconsistent output format. If you need the output in a specific structure, define it clearly. Use bullet points, numbered lists, or a template. The more explicit you are about output format, the more reliable the result.
Moving to Complex Tasks
Once you are comfortable with single-step prompts, try multi-turn tasks. For example, ask the model to generate a draft, then ask it to revise based on feedback. This teaches you how to manage dialogue history and maintain context across turns.
What to watch: in multi-turn conversations, the model might mix up information from earlier turns. Reiterate key instructions in each new turn. You can also ask the model to confirm its understanding before proceeding. This reduces errors and saves time.
Another complex task is asking the model to follow a chain of reasoning. For instance, “Analyse this customer complaint, identify the root cause, and suggest three solutions.” Break the reasoning into steps and ask for each step separately. Then combine the results. This approach mimics how you would teach a junior colleague.
Building Your Own Prompt Library
As you work through examples, save the prompts that work well. Organise them by task type: summarisation, generation, editing, data extraction, and so on. Over time, you will have a library of tested prompts you can reuse and adapt.
What to watch: a prompt that works for one model may not work for another. When you switch models, test your library again. Small differences in training data and architecture can change how a model interprets instructions. Update your prompts as needed.
This practice of building a personal toolkit is a skill that deepens with experience. Many professionals in Singapore attend AI prompting classes to accelerate this learning. The classes provide structured practice and feedback, which is harder to get when you work alone.
Applying Prompt Engineering to Your Work
The ultimate goal is to integrate prompt engineering into your daily workflow. Start with one task that you do repeatedly, such as drafting emails or creating reports. Use the steps above to refine the prompt until the output requires minimal editing. Then move on to another task.
What to watch: do not try to automate everything at once. Pick tasks where the cost of a mistake is low. As you gain confidence, tackle higher-stakes tasks. The same principles apply, but you may need more careful evaluation and testing.
Prompt engineering is a practical skill. The more real examples you work through, the better you become at spotting patterns and adjusting your approach. Whether you learn through self-study or through formal training, the key is consistent practice with real tasks.
For those in Singapore looking to develop this skill further, United Ceres College at 37 Middle Rd, #05-00, Singapore 188946 offers programmes that cover AI and related subjects. You can contact them at +65 8699 3609 for more information.