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?
Prompt engineering has drastically improved my output of work initiatives.
...
1 reply by Lynn Guimont
Feb 21, 2026 2:14 PM
Lynn Guimont
...
I am just getting started with AI. I used Claude to document a draft BRD for me, it generated a 12 page document within minutes but at the time I did not have the Mircosoft skill to be able to save it in Word. I was able to copy/paste it and it saved me a lot of time. I need to get better at the prompting because I did have to update a good portion of the draft document once I did have it in Word.
As a practicing CSEP & PMP, i have used AI in the bast to help expound on my ideas or in most cases streamline my documentation process for various tasks. I ran in to situations where i received very generalized or broad answers and sometimes was on the receiving end of AI hallucinations. Upon completion of this chapter, i created a spreadsheet with the template provided and i refined my prompts using the techniques provided within the course. My favorite is the CREATE along with chained prompts. I have seen a significant improvement with the accuracy of the responses. This will definitely improve my response validation time when it comes to reviewing responses from my models. Saving Changes...
Jennifer WalkerPrincipal Technical Program Manager| Sonos, Inc.Sharon, Ma, United States
Sometimes we think AI can do a lot more than it can do in one prompt. I recently used Claude AI to provide me a .csv file for validation before I had it load new work into Jira via a Jira MCP server. I asked for a .csv file to be created with project name, description and goals, and epic name with description and Must Have and Nice to Have requirements. This was all based on me uploading to Claude meeting transcripts and meeting notes for a new project along with best practices at our company on how to create Jira tickets. In the initial prompt, in addition to asking for the .csv file, I also asked Claude to compare the .csv file it generated with the Google Sheet I had manually created with the projects & epics so that I could validate Claude's output vs. my manual output. My request was too complex for one step! By breaking this into two steps and using a prompting chain, I was able to get the results I wanted - a .csv file from Claude and comparison of that file with the one I manually created.
Saving Changes...
Luke TimpeProject Consultant| Dyzana Consulting, LLC / CyberMassPhoenix, Az, United States
Dramatically, I have seen a lot of responses that miss things and the more responses and refinement will get there but if the models are training off your prompts it will get better over time. However, I don't like models training off my data due to proprietary items I know set me apart from others, so I ensure I get the responses and refine them outside of Models and input my own additional areas of refinement.
Prompt engineering has drastically improved my output of work initiatives.
I am just getting started with AI. I used Claude to document a draft BRD for me, it generated a 12 page document within minutes but at the time I did not have the Mircosoft skill to be able to save it in Word. I was able to copy/paste it and it saved me a lot of time. I need to get better at the prompting because I did have to update a good portion of the draft document once I did have it in Word. Saving Changes...
Monica BitlaComplaints Specialist| RAKBANKDubai, Dubai, United Arab Emirates
Carefully curating a prompt is more like an art which we develop and get better over time. It is essential that the prompt clearly specifies the role you want the AI/LLM to assume, the task it needs to perform and the format of the output you want. It is even better if the prompt is not overloaded with multiple tasks but rather a single task and thereby adapting the prompt chaining technique for AI to perform more tasks. Saving Changes...
Anonymous
Yes, but as I got better and the various models improved there has not been less emphasis on the prompts.
In my experience with GenAI, refining a prompt can completely change the quality of the output—often more than changing the tool itself.
Early on, I noticed that my first prompts usually produced technically correct but generic responses. The real shift happened when I started treating prompting as an iterative, inspect‑and‑adapt practice, rather than a one‑time request. Research and practitioner guidance consistently show that systematic prompt refinement—adding clarity, constraints, and feedback loops—leads to more accurate and relevant outputs.
What really stood out to me is that refinement isn’t about “gaming” the AI—it’s about making my own thinking clearer. Each iteration forces me to articulate assumptions, goals, and constraints more explicitly. Iterative prompting is widely recognized as a way to align AI outputs more closely with real‑world needs by introducing feedback and adjustment loops, much like how we improve work in Agile delivery. [ibm.com]
My biggest takeaway:
GenAI often gives you exactly what you ask for—just not always what you meant. Refining the prompt closes that gap.
In practice, this has shifted how I use AI: not as a one‑shot answer engine, but as a collaborative draft partner that improves as my clarity improves, thank you.
There are framewoks to create prompt. This is part of the Prompt Desing discipline. Those that gave me and the initiatives where I was included are:R-T-F (Role-Task-Format), T-A-G (Task, action, goal), B-A-B (Before, after, bridge), C-A-R-E (context, action, result, example), R-I-S-E (role, input, steps, expectations).
PMI Team Member
