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?
A prompt needs to be refined accordingly for the best output. For large questions, its always better to use the Chain prompting mechanism for the desired and actionable output. Saving Changes...
Anonymous
It makes a world of difference. There is no longer large uncertainty with credibility after refining the GenAI. That gap of reliability becomes smaller. You also expand the brain of the AI as you refine and give feedback. Saving Changes...
In my experience, refining a prompt can drastically change the quality of the AI output when the first request is too broad.
In a retail technology and project delivery environment, I may ask AI to help prepare a project update for a POS rollout or system integration activity. If the prompt simply says, “prepare a project status update,” the response may look polished, but it is usually too generic. It may not separate risks from issues, may not reflect vendor dependencies, may not highlight store readiness, and may not give the right level of detail for leadership or operations teams.
The output improves when I refine the prompt step by step. For example, I would first ask AI to summarize the project situation, then ask it to identify key risks, then separate open issues from risks, then prepare stakeholder-specific updates, and finally convert the result into an executive summary.
The specific changes that make the biggest difference are adding business context, defining the audience, clarifying the expected format, and asking AI to focus on one or two tasks at a time. I also find it useful to tell the AI what not to do, such as not making assumptions about dates, costs, approvals, or technical readiness unless those details are provided.
For me, prompt refinement is not only about getting a better-looking answer. It is about making the output more usable for real project work. The first response is usually a draft. The refined response becomes closer to something a project manager can validate, adjust, and use for decision-making or stakeholder communication.
So the biggest lesson for me is that AI output quality improves when the project manager brings context, structure, and judgment into the conversation. Saving Changes...
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).
Thank you, Sergio. I recently joined PMI, and I am still learning from the community discussions, so your sharing on prompt frameworks is very useful.
I agree that prompt frameworks are helpful because they give structure to how we interact with AI, especially when the task is related to project delivery.
In my own retail technology and project management work, I see these frameworks being useful in different situations:
RTF - Role, Task, Format I use this for quick and direct outputs. For example: “Act as a project manager. Prepare a weekly POS rollout status update. Present it in bullet points with progress, risks, issues, and next steps.”
TAG - Task, Action, Goal This is useful when I want the AI to focus on the intended outcome. For example: “Review the open project issues, group them by action owner, and help the team focus on items that may affect store readiness.”
BAB - Before, After, Bridge This works well for explaining transformation or change. For example: “Before, the process was manual and inconsistent. After, the process should be standardized and visible. Bridge the gap by proposing a practical implementation approach.”
CARE - Context, Action, Result, Example I find this useful when preparing communication or stakeholder updates. For example: “Given the current retail system migration context, draft a message to business users explaining the required action, expected result, and provide a simple example of what they need to check.”
RISE - Role, Input, Steps, Expectations This is helpful for more structured project work. For example: “Act as a PMO reviewer. Use the project issue log as input. Review it step by step and highlight missing owners, unclear deadlines, and items requiring escalation.”
For me, the framework is not the final answer. It is a starting structure. The real value comes when the project manager adds business context, constraints, stakeholder expectations, and then validates whether the AI output is accurate and useful for the actual project situation. Saving Changes...
The iterative refinement always works to get the desired output.
Saving Changes...
Mohamed NaamaProject Management| Alkhorayef Water and Power TechnologyAl-Arish, SIN, Egypt
h2/h2
Refining a prompt can significantly improve the quality of GenAI outputs. Adding context, defining a clear role, or specifying the objective often shifts the response from generic to precise and more aligned with project needs. In many cases, the prompt itself becomes the main factor that determines how accurate and useful the final output will be.
Saving Changes...
Mohamed NaamaProject Management| Alkhorayef Water and Power TechnologyAl-Arish, SIN, Egypt
Refining a prompt can significantly improve the quality of GenAI outputs. Adding context, defining a clear role, or specifying the objective often shifts the response from generic to precise and more aligned with project needs. In many cases, the prompt itself becomes the main factor that determines how accurate and useful the final output will be.
Saving Changes...
Archana DengiTata communications LimitedGulbarga, KA, India
As AI uses large data set to process our prompt, basis first prompt it understands what we try to get from it but not exactly able to deliver. By Refining prompts users understanding and response understanding is getting better to achieve the targeted output. RTF will definitely help here. Saving Changes...
Samir GoswamiSenior Project Management| Thomson ReutersBengaluru, KA, India
Specificity and clarity in the prompts are the key to have a better output.
PMI Team Member
