Director, Learning Design & Development| PMIAsheville, NC, United States
With Generative AI, iteratively refining and optimizing prompts can lead to better AI-generated results. This may involve adjusting the specificity or clarity of the prompt to increase relevance and accuracy of results.
What examples do you have of how improving a prompt drastically changed the output quality? What specific changes did you make that led to the improvement?
Refining prompts drastically improve the quality of the output. I tried many times with conversational style without repeating what I asked earlier. It understands well and remembers the previous discussion and customize the results accordingly.
AI is super smart to understand whatever we say but the way how human understand, the same way it understands. Saving Changes...
Vasco MarquesIT Manager| lcs valanticLisboa, Portugal
In my personal experience with a private GenAI based on ChatGPT, i found that using a well structured Prompt Prefix (built with some iterations) I could get a much better response namely in the output format even using very simple prompts. I also like very much the idea/ concept of building and refining prompts as if you're talking with another human or group of humans about some complex process or concept, the more precise you frame and provide context about what you're discussing and/ or sharing, the more precise you are able to describe the elements, their relationships and the rules that govern what you want to convey to the other person or group, the level of understanding that they get increases, i.e. the response to your prompt is better, the more open to feedback, active listening and observing their reactions, the more you get the chance to rephrase, reframe your concept and greater is the chance to increase the understanding of the concept or the idea that you're trying to convey, i.e. the more you are willing to use a prompt response as feedback on your initial prompt design, the better chances you get a better reponse upon refining the initial prompt Saving Changes...
I think that not only ensuring clarity, providing context but also iteratively testing, avoiding bias and encouraging creativity we can get better ouput quality Saving Changes...
An example is when I had asked AI to generate a visual process flow diagram and it first gave me a diagram where step 1 was on the right bottom side of the diagram and the process visually went left, down, swooped into a circle, then diagonally up and left the last step. I iterated to change the shapes from filled in blue circles to rectangles, as well as provide information that it should begin at the top for step one and as each step number increased it should flow directly down. I ended up with a human-readable diagram that followed a top-down visual process. Saving Changes...
Yes, it improves the output and gives desired results. Saving Changes...
Christine PurvianceSr. HR Manager| Avista UtilitiesCheney, Wa, United States
I'm not part of the AI pilot team at my company. But I did use a refining technique, actually led by the AI, to help me get to what I was wanting for a personal use. Saving Changes...
Anonymous
Giving more context and being more specific with regards to the expected output has improved initial responses in my experience.
The more you use it, the more you "train on the job", the better the prompts. Of course following guidelines like CREATE are really useful, as they help maintain the focus when facing challenging prompts. Saving Changes...
Kelly WelchSenior Project Management| City of Austin, TX
In my experience with GenAI, refining a prompt can completely transform the quality and relevance of the output. Early on, I’d often enter vague or broad prompts and receive generic responses that didn’t quite hit the mark. But once I started being more intentional—adding context, specifying tone, and clearly stating the goal—the difference was night and day.
For example, asking “Help me write a professional email” might produce a decent draft, but something like “Help me draft a concise, polite email to a coworker explaining a missed deadline due to technical issues, and propose a new delivery date” results in a much more tailored and useful message.
I've also found that iterating—refining the prompt after seeing the first draft—can help narrow in on exactly what I need. It’s less about “getting it right the first time” and more about treating the process like a collaboration. The clearer I am, the better GenAI performs.
I found that really taking time to define in a concise way the character of the person that would be an expert on the topic that I am asking the AI system, has helped to positively impact the quality of the response. Also asking for the response broken down into specific categories helps to provide better direction on exactly what I'm expecting. Saving Changes...
PMI Team Member
