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?
In my experience using GenAI, I’ve found that establishing clear objectives is crucial to ensure the outputs align with my goals. Knowing exactly what I want to achieve helps in setting precise criteria for success. GenAI significantly speeds up the process of idea gathering and structuring, making it an invaluable tool for efficiency.
However, it’s important to remember that AI doesn’t replace human expertise and judgment. Implementing strong testing protocols and continuously refining the AI model based on feedback are essential practices.
Additionally, cross-verifying AI outputs with human expertise ensures accuracy and relevance. In my view, these steps collectively help in validating and checking the outputs effectively.
Best regards,
Daniel S.
Saving Changes...
Tareq A. Al BehairiProject Management Consultant| Independent Consultants and TrainerGcc, Kuwait
I believe the most effective way to validate Gen AI outputs is to engage human input. Project managers should not take the AI-generated output as the ultimate outcome but should act as judges to investigate the results through a human-in-the-loop approach or by utilizing external experts for sanity checks. For example, if an AI model suggests a particular project timeline, the project manager should review this recommendation, considering their own experience and consulting with team members to ensure its feasibility. Additionally, crafting good prompts is essential for generating accurate and useful answers. For instance, instead of asking a vague question like "What should I do next?", a more specific prompt such as "What are the best practices for managing a project timeline in the renewable energy sector?" will yield more actionable insights. This combination of human oversight and well-crafted prompts ensures that AI serves as a valuable tool rather than a definitive authority.
Remark : This response is powered by OpenAI's GPT
Prompt :
( Refine the post below ,in one cohesive paragraph .i want to post it to PMI community discussion." i believe the most validation way for the Gen AI outputs is to engage the human input .do not take the output as ultimate outcome ,PM should be as judge to investigate the result (human in loop) or utilize external experts to make sanity checks. All of pervious besides good prompts generate good answers )
The more specific your questions or instructions, the more likely the AI will provide useful answers. Avoid vague or overly broad queries.
When possible, provide context for your request. Background information, relevant details, or examples can help the AI better understand and fulfill your needs. Saving Changes...
Moise LontsiProjects & Program Management| Point Hope MaritimeVictoria, Canada
We enhance the response's accuracy by employing the "CREATE" approach, and iteratively refining the prompt. Saving Changes...
The art of getting an output using iterative refinement is key to obtaining a more accurate output. This is important especially where the Requirement demands a long Prompt to be Composed. The iteration allows human validation at each step as the output is built.
Asking AI to help build a Prompt, in the hope of getting an accurate output is probably not the best way. The human being simply has to be very clear on what one needs to achieve Saving Changes...
Providing the AI ​​with the appropriate context for our company or particular situation, through examples or documentation related to the case. Then iterating on aspects that will provide more precision and refinement to the answers, until reaching an answer that contains the desired objectives and results. Saving Changes...
Continuous refinement, much like an agile development approach works has worked for me. Typically I can get to some reasonably good content in a fair amount of time and then put the finishing touches on from there. Saving Changes...
Jordon WeberPM I| Kiewit Infrastructure Engineers Durham, NC, United States
Get cozy with your project goals. Then, make sure your data is top-notch—no room for junk! Team up with domain experts, stay in the loop on AI trends, and give your results a reality check. Oh, and don’t forget the basics: understand the tasks you throw at Gen AI. It’s like teaching a dog new tricks—know what you’re asking for! Saving Changes...
Anonymous
Have not tried GenAI/prompt engineering on production systems. However, the basic steps (understand, design, test, implement) should be performed before any full-scale adoption of the AI systems. Saving Changes...
"If they have moving sidewalks in the future, when you get on them, I think you should have to assume sort of a walking shape so as not to frighten the dogs."
PMI Team Member
