Generative AI has come to make our lives much
easier by swiftly producing texts, code, or research information that would
otherwise take longer to obtain. But there is a downside, though. The system
needs to be properly guided.
It works more like the popular computer term,
“garbage in, garbage out”. This simply means what you give is what you get. If
the quality of your probe is poor or vague, the response of the system will
also likely be poor.
To get the best out of generative AI, you may
want to follow the steps below. It’s a bit lengthy but once you get used to it,
it becomes a piece of cake.
Step 1: Properly
state what you need
This part is very important, state what you
want in very simple terms. This stage is the beginning and it needs
specificity. Directly state your purpose and expectation. It is also important
to mention your audience and the format that you expect. This will help the
system to tailor its response to your need.
Avoid short or vague instructions like “write
something about climate change.” Instead, say:
- “Write a 600-word article about climate change impacts on
agriculture.” - “Use a neutral tone and include subheadings.”
A clear goal tells the AI exactly what success
looks like, saving you time later.
Step 2:
Provide context and background
AI performs best when it understands the
situation. Give it background details, such as product descriptions, industry
focus, or previous examples. For instance:
- “Write a press-style summary for the Huawei Mate 70 Pro using these
specs.” - “Base the report on data from Q2 2025.”
Context helps the model stay accurate and
prevents generic or irrelevant output.
Step 3: Pick the appropriate model that fits your needs
Don’t just jump on a model because its free and then lament when its output
is poor. Different models are designed for different needs. Those who just want
to write will not use the same model as coders.
Using the wrong model can limit your outcome. Here are some examples of generative AI types and
what they do best:• Text
models – ChatGPT, Gemini, Claude
• Image
models – DALL·E, Midjourney, Stable Diffusion
• Code models – GitHub Copilot, Replit AI
• Audio
models – Suno, Udio, MusicLM
• Video models – Runway, Pika, Synthesia
• 3D or design models – Spline AI,
Leonardo
Choosing the right model ensures your results are relevant, polished, and
tailored to your specific goal.
Step 4:
Write detailed prompts
Good results come from good prompts. Always
give complete instructions that cover:
- The task: What you want it to do (e.g., write,
summarize, explain). - The format:
Length, tone, or structure. - The examples: What
style or outcome you prefer.
Example: “Write a 500-word neutral article
about Samsung’s Galaxy Tab S10. Include display size, dimensions, and specs.
Avoid promotional language.”
This structure gives the model a clear
blueprint to follow.
Step 5:
Break down complex tasks
If your request involves multiple parts, don’t
do everything in one prompt. Divide it into smaller stages. For example:
- Ask for an outline first.
- Then request the introduction.
- Next, expand each section.
- Finally, ask it to polish the entire text.
Breaking tasks this way keeps the AI focused
and ensures a smoother, higher-quality final result.
Step 6: Avoid
very long texts and ask for multiple versions
Personally, I would recommend that the maximum
number of words you request is 1000 words, preferably 500 words. The longer
the text you require generative AI to write, the more likely it will pad your
text with “rubbish”.
If your text is short, less than 400, then ask
for multiple versions. For example, your probe could include “Produce three
options for this text, and each option should use a different tone.”
You can compare the options and pick the
one that best appeals to you.
Step 7:
Review and refine the output
Treat the first result as a draft, not the
final version. Read carefully and ask the AI to revise specific areas:
- “Make the tone more formal.”
- “Add more technical details about performance.”
- “Shorten the conclusion without losing meaning.”
Refining through follow-up prompts turns rough
drafts into polished work.
Step 8:
Verify information and sources
AI systems are very far from perfect. If you
use it blindly, it will completely embarrass you. A lawyer in the US was
recently fined $10,000 because he blindly used ChatGPT to prepare his case, and
the system included legal citations that do not exist.
Therefore, there is a need to check and double-check whatever AI gives to you. Check the facts, data, and citations. Please,
if any name is mentioned by the system, confirm the name. Even when you ask AI
to produce references, it may produce fake references. Thus, verification is
very important.
Step 9:
Edit with human judgment
After swiftly writing the basics with AI, it
is important to humanize the text with your own human judgment. In this stage,
you will revise your text based on your findings. You may need to adjust the
flow and maybe tone. In many cases, if you do this properly, the readability will
be better. Ensure that you do the following
- Replace robotic phrasing with natural expressions.
- Remove redundancy or filler sentences.
Human editing ensures the final result sounds
intentional and authentic.
Step 10:
Save and reuse effective prompts
When you find a prompt that gives great
results, save it. Create a small library of prompt templates for your common
tasks. For example:
- “Article summary prompt.”
- “Technical explanation prompt.”
- “Data analysis prompt.”
Using templates makes your workflow faster and
ensures consistent quality over time.
Final Word
While using generative AI, I recommend that
you maintain good ethics and protect privacy. Do not share the personal
information of somebody or company-sensitive information because AI gave it
to you.
