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?
Jaswinder LambaFunctional Manager| MicrosoftSammamish, Wa, United States
Keeping AI prompts clear and concise significantly improves the quality of results, as explained in the session, we can break a large problem into smaller manageable tasks and make it easier to obtain accurate and relevant outputs aligned with the goals of the project. Saving Changes...
Pablo SerraProject Manager| H&T PresspartAltafulla, Tarragona, Spain
you see a clear difference between vague answers to more specific outputs Saving Changes...
I have had little experience using GenAI such as ChatGPT, but the few times I have done so, I did so without knowing how to properly structure the prompts to obtain better responses.
I think that, intuitively, I sometimes made improvement questions based on some things I read or that I managed to find on the Internet, but without having a structure or a defined way to do it, such as the RTF or the CREATE technique that we saw in this course. I did notice, when making changes or refining the prompt, that I obtained better results in the adjusted response.
I cannot imagine the change I would achieve if I mastered Prompt Engineering.
In my experience leading HubSpot implementations globally, refining prompts in GenAI is like fine-tuning a project brief, it changes everything. An initial, vague prompt might yield generic output, but layering in context, expected tone, and formatting transforms the results into something actionable and client-ready. I’ve seen outputs shift from surface-level summaries to strategic, insight-driven content just by restructuring the prompt to mirror stakeholder expectations. Prompt iteration feels a lot like agile sprints: define, test, refine. Sometimes, what unlocks quality is as simple as stating the role the AI should adopt or providing a clear use case. It’s not just about getting any answer, it’s about getting the right one. Prompt engineering has become an essential skill for maximizing the value GenAI brings to project management and client communication. Saving Changes...
It had certainly enhanced the quality and minimized the number of iterations to get to the needed output. Saving Changes...
Anonymous
Here's a more detailed look at how refining prompts impacts GenAI output:
1. Increased Relevance and Specificity:
Moving Beyond General Prompts:
Instead of vague requests like "write a story," a refined prompt could specify a genre (e.g., "write a science fiction story about a robot uprising"), a tone (e.g., "write a humorous short story"), or a specific theme (e.g., "write a story about the impact of social media on teenagers"). This leads to more tailored and relevant results.
Adding Context and Constraints:
Refining prompts can include information about the context of the request, such as the target audience, the length of the output, or even specific keywords or phrases to incorporate.
Using Examples:
Providing the AI with examples of the desired output type can help it understand the desired format and style. For example, you could include a sample paragraph or sentence that demonstrates the desired tone or vocabulary.
2. Improved Clarity and Accuracy:
Avoiding Ambiguity:
A well-refined prompt will clearly state what you want the AI to do, minimizing ambiguity and reducing the risk of the AI misunderstanding your request. For example, instead of saying "summarize this text," a refined prompt might say "summarize the main arguments of this text in 2 sentences".
Using Precise Language:
Refining prompts involves selecting the most accurate and specific vocabulary to describe what you want. This ensures the AI understands the nuances of your request and produces more precise output.
Ensuring Logical Flow:
A well-structured prompt can guide the AI through a logical sequence of steps or ideas, resulting in a more coherent and organized output.
3. Enhanced Control and Specificity:
Role Prompting:
Assigning a specific role or persona to the AI (e.g., "act as a seasoned journalist," "be a creative writer") can significantly influence the style and tone of the output.
Task-Specific Instructions:
Providing detailed instructions about the specific task you want the AI to perform (e.g., "generate a list of keywords for a blog post," "write a marketing copy for a product launch") can lead to more focused and relevant results.
Iterative Refinement:
GenAI output is often an iterative process. Refining the prompt based on the initial output allows you to guide the AI toward a more perfect result, improving both relevance and accuracy. Saving Changes...
This morning, one of my LinkedIn contacts complained that GenAI tools don't seem to have the ability to craft decent quality PMP practice exam questions. He had used the public version of ChatGPT. I decided to check the same with PMI Infinity and got better results - seven out of ten questions were acceptable.
My first prompt was "Generate ten different questions about project management which would be similar in style and level of difficulty to what is asked on the PMP exam"
It only gave me the questions but neglected to provide any answers. Realizing that this was likely it interpreting what I had asked for "as is", I then added: "Generate ten different questions about project management which would be similar in style and level of difficulty to what is asked on the PMP exam".
With that it was able to provided more useful output.
Kiron
Thank you for posting. Saving Changes...
Mary KrebsDirector of Research & Strategy| Brightline Stategies LLC
Jun 21, 2024 10:50 AM
Replying to Laura Lazzerini
...
I think it is important to give the context and also to refine, asking for a different output in case that the first one is not completely suitable to our purpose or to the outcome that we were looking for. I think that consistency and preseverance in looking for the result, is crucial as well.
Agreed. Focusing on a solid persona alongside an example of desired output also seems to generate higher quality responses. Saving Changes...
This morning, one of my LinkedIn contacts complained that GenAI tools don't seem to have the ability to craft decent quality PMP practice exam questions. He had used the public version of ChatGPT. I decided to check the same with PMI Infinity and got better results - seven out of ten questions were acceptable.
My first prompt was "Generate ten different questions about project management which would be similar in style and level of difficulty to what is asked on the PMP exam"
It only gave me the questions but neglected to provide any answers. Realizing that this was likely it interpreting what I had asked for "as is", I then added: "Generate ten different questions about project management which would be similar in style and level of difficulty to what is asked on the PMP exam".
With that it was able to provided more useful output.
Kiron
I like the idea of prompting the AI to provide a step by step answer to questions as it provides context and rationale! Saving Changes...
PMI Team Member
