Disclaimer: I’m not an AI engineer or a data scientist. But I’m someone who has taken up courses on prompt engineering and experimented a lot with prompts. If you are an AI professional, feel free to correct me.
70% of the people while using a AI tool, just type in a question and hope for magic to happen. I also did that for months. Don’t get me wrong the answers were fine, but they did lack depth, originality, or focus. Then one day I took a course on prompt engineering- I’ve listed my findings in Beginner’s Guide to ChatGPT Prompt Engineering for Better AI Responses, which helped me start on the right foot. From then, I’ve learnt, read and experimented more, and I’ve come up with my list of AI prompt wins, which I’ll be sharing with you in this article.
Did you know that most of the times AI frustrations aren’t caused by the tool, but they are actually caused by the way we ask for things? So when you give an AI model vague, overloaded, or biased instructions, it is going to struggle to give you the depth, clarity, and creativity you want. For getting your desired output, incorporate the following prompt wins and thank me later. Let’s begin!
1. Be Specific and Scoped
Why It Works:
AI models have no common sense. So you can’t assume it’ll guess or know things. Specificity gives AI a precise target and reduces the scope of guesswork. What happens when you define a scope is that it ensures the model stays focused on your intended outcome, without drifting into unrelated tangents. And what happens when your prompt is very broad? AI will will give you a generic overview instead of a focused and a tailored answer.
Example Win:
“Act as a tech journalist. Write a 500-word blog post explaining how small businesses can use AI for customer service, with 3 concrete examples.”
What Not to Prompt:
“Write me a blog post about AI.” – This is too vague and not specific at all. At the same time it does not give a clear direction for tone, audience, or focus.
2. Break Big Tasks Into Smaller Ones (Prompt Decomposition)
Why It Works:
When working with AI models it is important to understand how LLMs work. It’s a fact that LLMs do best when they focus on one task at a time. So if you ask for a huge task, there’s a likelihood that the AI model will rush into the task and generate not so desired result. So let’s say you ask for a research report, a slide deck, and a polished article in one go, it might rush parts of the request, leading to thin or inconsistent results.
Example Win:
- “Research the top 5 AI ethics challenges.”
- “Summarize those findings in 300 words.”
- “Turn that summary into 5 slide bullet points.”
It’s always advisable to break huge tasks into bite sized tasks, to get the desired result.
What Not to Prompt:
“Research AI ethics, summarize it, and make slides.” – This kind of prompt usually overloads the model and splits its focus.
3. Assign a Role or Perspective
Why It Works:
AI loves hypothetical scenarios. It works best when asked to roleplay. Giving the AI a role sets the style, voice, and expertise level for the answer. If one doesn’t do this AI is going to respond in generic and one-size-fits-all way.
Example Win:
“You’re a high school teacher. Explain blockchain to 15-year-olds using a school library analogy.”
What Not to Prompt:
“Explain blockchain.” – This kind of prompt, doesn’t specify your level of understanding/ awareness of the concept, as well as it doesn’t specify the audience, tone, or angle.
4. Encourage Step-by-Step Reasoning (Train-of-Thought)
Why It Works:
When asked for reasoning, it improves the accuracy as well as transparency. What this does is it helps AI avoid skipping critical thinking steps and reduces the risk of AI hallucinations.
Example Win:
“Think step-by-step. Show your reasoning before giving the final answer to this logic puzzle.”
What Not to Prompt:
“Solve this puzzle.” – This kind of prompt invites a rushed, possibly incorrect answer with no explanation.
5. Frame Prompts to Challenge Assumptions
Why It Works:
You should know by now that models like ChatGPT are always agreeable. It’s true that AI models tend to agree with you by default. So let’s say your prompt is leading or biased, an AI model may simply reinforce your opinion instead of offering an correct, objective answer.
Example Win:
“Is my assumption correct, or is there a better explanation? Explain why.”
What Not to Prompt:
“I think X is true, right?” – What this does is it signals the AI model that you want agreement rather than truth.
6. Provide Context Before the Task (Priming)
Why It Works:
When you provide AI with context, it primes it to work within your intended boundaries. What happens when you don’t context- the AI fills in gaps with assumptions that may not match your needs.
Example Win:
“Here’s my draft email to a potential investor. Make it concise and persuasive, keeping a professional but friendly tone.”
What Not to Prompt:
“Improve my email.” – This kind of prompt again has no indication of audience, tone, or purpose.
7. Confirm Shared Understanding Before Proceeding
Why It Works:
This actually was a great tip that I learnt, when you check alignment, that is to see if you’re at the same page, right before the task begins, it prevents wasted effort and the need to prompt again. So this way you can catch and correct the AI if it has misunderstood your task.
Example Win:
“Here’s my request: ____. Can you repeat this back to me in your own words before starting?”
What Not to do:
Jumping straight into the task without asking the AI to confirm if it understands your instructions or not.
8. Iterate and Refine
Why It Works:
The hard truth is that the first drafts are rarely perfect or completely accurate. Don’t treat AI as the sole tool, do your research first and treat AI as a collaborator. When one focuses on refining and improving outputs in steps, they produce the best results from their prompts.
Example Win:
“This is good. Now rewrite it with a more persuasive opening and add two real-world examples.”
What Not to do:
Accepting the first response without providing it any feedback or asking it for further refinement.
9. Use Cross-Domain Creativity
Why It Works:
AI actually excels at blending concepts from unrelated fields. It uses its cross domain creativity and produces fresh perspectives and memorable explanations.
Example Win:
“Explain machine learning like a chef teaching a recipe.”
What Not to do:
Ask an AI model for a plain definition without inviting creative reframing.
Bottom Line
Think of your AI like a talented teammate who can do almost anything, when you tell them exactly what you want. The clearer your direction, the better the performance. If you’re curious about how prompts themselves are turning into a product, you’ll enjoy reading about AI prompts being sold online and why people are buying them.
