Director, Learning Design & Development| PMIAsheville, NC, United States
Validating and checking outputs is critical when working with AI systems like Generative AI. Such validation approaches may include establishing clear criteria, implementing strong testing protocols, and continuous refinement.
In your experience with AI, what are some best practices for ensuring the results you receive are accurate, relevant, and aligned with your original goals?
Be specific in your questions to AI to generate the correct output and compare to original goals. Modify prompts as needed to adjust and reset to goal. Then validate and assess outputs until satisified - ultimately it is your choice. Saving Changes...
Eric LaylandConsultant| MattersSeattle, WA, United States
I come from a marketing background and developing personas and identifying the target audience comes naturally. I've found that defining the persona and supplying details about the intended audience greatly influences the output. I will also ask the LLM, "Is there anything I'm missing or that is unclear? If so, please ask me additional questions to provide the level of clarity sought." Initially I would write fairly long prompts of a paragraph, sometimes more. Now I try to break up the prompt into discrete "units of output" I'm seeking. After setting the stage with the personas, intended audience and answering clarifying questions, I'll create a string of prompts that addresses each step towards the desired level of detail in the outputs. Saving Changes...
Aaron GosselinLighting Project Manager| LineWorks EngineeringBrandon, Fl, United States
Many people in the community have hit on it already but I believe it is using clear and concise prompts, reading through the LLM's response, then modifying the prompt if necessary to create a pattern of responses that are more in line with a usable model output. Saving Changes...
Vinayak DeoSr. Project Manager| FedExCollierville, Tn, United States
Since I am an IT Professional, I look at it from IT project management perspective, and in order to validate AI generated outputs in IT projects, we usually follow a five-step review framework. First, define expectations by setting objectives, scope, and evaluation criteria to ensure alignment with project goals. Next, conduct an initial review to check for general accuracy, completeness, and coherence. Then, perform deep validation by cross-referencing the output with system data, reports, and SME inputs to confirm factual correctness. Assess bias and ethics by ensuring neutrality, fairness, and balanced representation. Finally, refine and approve the output through human-in-the-loop oversight, addressing errors and confirming it meets project needs before use. Saving Changes...
Some of my items may be redundant but the most important things in my experience so far is:
Be precise and clear.
Be sure you explain jargon or specialized terminology
Provide the context for all of your requests
Be sure you provide the outcomes you are expecting
Experiment and refine as you go
I've found breaking down big problems can be better refined by chunking the whole into natural sections and working to refine each section and then working to put them back together.
I agree that you have to clearly define the objective of the prompt and avoid ambiguity—be as specific as possible. Also, provide relevant background information and examples. It is always good to follow the role, task and format method. It will help you produce accurate, reliable, and relevant results. Saving Changes...
Anonymous
Enhanced Context and Evaluation/Validation Criterias are a combination to pursue the desired accuracy level. Saving Changes...
George Wardell, PMPIT Project Manager| PMI Atlanta ChapterKennesaw, GA, United States
There may not be a way to get perfect results, but you have to be clear and concise when interacting with the AI. I think giving examples may be the key to getting the results that you want and also being very knowledgeable in the subject matter so you can identify a hallucination. Saving Changes...
Continuously validate outputs by cross-referencing with reliable sources and applying human judgment to detect biases or errors. Also, regularly refine inputs based on feedback and modify AI prompts to improve relevance Saving Changes...
I follow RTF and then start feedback loops. If I'm still not satisfied with the responses then I reverse the roles and provide specific answers to the questions raised by AI for tailoring the solutions as per my need. Saving Changes...
Jumarin IsmailDirector Advisory & Forensic Services – Infrastructure & Capital ProjectRiyadh, Saudi Arabia
Thanks for sharing the thought,
The outlined best practices are spot-on, especially regarding clear objectives and verifying AI-generated outputs. One key aspect I would like emphasize is the importance of continuously refining prompts AI responses improve significantly when we provide more context and iterative feedback. Additionally, while AI enhances efficiency, human oversight remains essential to filter out biases and ensure strategic alignment. Overall, leveraging AI as a tool rather than a replacement is crucial for maximizing its value. Great insights!" Saving Changes...
PMI Team Member
