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?
Ivan KozhevnikovCEO | Strategic Business Leader with Consulting ExpertiseAlmaty, Kazakhstan
In my experience, GenAI can be extremely helpful - but it doesn’t guarantee accuracy and is prone to generating confident but incorrect results, a known issue often referred to as “hallucination”. That’s why validation is essential. I often ask for sources, assess their relevance based on my own expertise, or use guiding questions like “are you sure?” - surprisingly, this often leads to more refined or cautious responses. Still, without a clear internal vision of the expected outcome, relying fully on GenAI can be risky. Especially in data-heavy tasks like spreadsheet analysis, where numbers may be misinterpreted or incorrectly aggregated.
One thing I’ve learned: you need to know what “good” looks like - otherwise, you might validate a convincing hallucination.
Saving Changes...
HAMADA BADRProject and planning Manager engineer| ACTCKuwait, Egypt
you should be specific and use method such CREAT Saving Changes...
It is a great question. Even though AI won't give us an accurate response at the beginning it is important to provide detailed, precise and clear information. Try to use simple terms and everything relies on iteration because if you think you didn't receive a good response, try again , be more specific, gage, check, until you get the response that you want. Saving Changes...
Jaswinder LambaFunctional Manager| MicrosoftSammamish, Wa, United States
To ensure the results I receive from the AI are accurate, relevant, and aligned with the original goals, I follow a few key best practices. First, I provide clear and detailed prompts that include context, learning objectives, and desired outcomes. This helps guide the AI’s responses in the right direction. To effectively manage training development projects, I validate the outputs against the course outline and organizational goals to ensure alignment. Iteration is important—I review, refine, and adjust the content as needed. Finally, I combine AI-generated content with subject matter expertise to ensure both accuracy and instructional integrity. Saving Changes...
After a good and refined prompts, we need to validate the information given by GenAI with the team. From my point of view, this is the unique way to validate the information is accurate, relevant and aligned with the original goals. Saving Changes...
Pablo SerraProject Manager| H&T PresspartAltafulla, Tarragona, Spain
Ask the AI about the information sources and also based on what hthe AI considers each answer is aligned with each objective. Saving Changes...
As AI continues to evolve, learning to speak the language of AI and how best to get the output we are looking for will be a continual process.
As a PM with experience implementing AI-powered tools like HubSpot globally, I’ve learned that AI’s value is only as strong as the strategy behind its use. Ensuring accurate, relevant results begins with clear, structured prompts—but also demands a human-in-the-loop mindset. I treat AI interactions like stakeholder conversations: the better the context, the better the output. I also apply iterative testing, much like building a complex dashboard or workflow—validate in stages, refine continuously. Cross-checking AI-generated insights with trusted data sources or SMEs is key. Finally, I align every output against the original goal to avoid "drift." In the end, prompt engineering isn’t about replacing expertise—it’s about amplifying it with thoughtful design, validation, and alignment. AI doesn’t replace PM intuition—it sharpens it when used deliberately. Saving Changes...
Leah LavalleeProgram Manager| AtkinsSc, United States
You need to define your objectives clearly and include any qualifying or background information. This will minimize vague and unrelevvant responses. Saving Changes...
Anonymous
One may use the CREATE method and attach documents that are specific to one's projects or lessons learned so the LLM has the latest information. Saving Changes...
Anonymous
As discussed throughout the course - AI in its current state is just a tool. The validation can be done by AI as part of an iterative process, but ultimately human expertise is really needed to validate the output.
In the context of project management, AI can help you build the documentation -whether it be the scope document, the stakeholder matrix, the WBS, etc. However, it takes the PM expertise to identify the need for these documents, the oversight to ensure they are created, the human interaction to ensure validation. The scope document needs to be evaluated by the customers and sponsors. The stakeholder matrix needs to be confirmed by all project team, and the WBS needs to be validated by the teams performing the work. They themselves might also use AI to evaluate these tools and perhaps as the tools get more sophisticated, human interaction may be reduced. Saving Changes...
PMI Team Member
