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?
Stephanie MillerProgram Manager| Port Authority of NY and NJRichmond, VA, United States
Use the CREATE method to provide the original inputs and refine the prompts with new information, or a new version of the question (i.e. flip the approach) but refer to the original inputs to maintain the goals and alignment of purpose with the original prompt
Saving Changes...
Stephanie MillerProgram Manager| Port Authority of NY and NJRichmond, VA, United States
Use the CREATE method to provide the original inputs and refine the prompts with new information, or a new version of the question (i.e. flip the approach) but refer to the original inputs to maintain the goals and alignment of purpose with the original prompt
Saving Changes...
Hellen charlessseo expert| Digital MarketingHouston, United States
One of the most effective practices is to treat AI as a starting point, not the final authority. I try to verify important information against official documentation or trusted sources, break complex tasks into smaller prompts, and test outputs with real examples. Providing clear context and refining prompts based on the results also improves accuracy. A quick review before using AI-generated content can catch mistakes and ensure it aligns with the original goal. Saving Changes...
Hellen charlessseo expert| Digital MarketingHouston, United States
One of the most effective practices is to treat AI as a starting point, not the final authority. I try to verify important information against official documentation or trusted sources, break complex tasks into smaller prompts, and test outputs with real examples. Providing clear context and refining prompts based on the results also improves accuracy. A quick review before using AI-generated content can catch mistakes and ensure it aligns with the original goal. Saving Changes...
Dheeraj PalProgram Manager| KyndrylDelhi, Delhi, India
First your objective must be very clear - What is your requirement, what you want from LLM to provide you, around that you need to provide clear detailed information to LLM so that it will help you to achieve your requirement. Verification is must before scaling up.
Saving Changes...
José WRIGHTProject Management| NOVA SWISSParis, IDF, France
With my small experience so far on the topic, I would say that verifiyng and comparing AI responses to other experts documentation is a must. It is also a good practice to systematically ask the AI for its sources and ask to elaborate the anwser, using refinement inputs. Saving Changes...
style.ql-indent-1 { margin-left: 3em; }/style1. Define clear objectives and provide context
Clearly explain what you want the AI to achieve.
Provide relevant background information, constraints, audience, and desired format.
Example: Instead of asking “Create a project plan”, specify “Create a 6-month solar farm project execution plan including milestones, risks, resources, and regulatory approvals.”
✅ 2. Use high-quality and relevant data
Ensure the information provided to the AI is accurate, complete, and up to date.
Avoid relying on incomplete, biased, or outdated data.
✅ 3. Write effective prompts
Be specific about:
Role (e.g., “Act as a PMP-certified project manager”)
Task
Context
Expected output format
Success criteria
✅ 4. Validate AI-generated outputs
Review results against:
Business objectives
Project requirements
Industry standards
Compliance and policies
Do not assume AI outputs are always correct.
✅ 5. Apply human judgment (Human-in-the-loop)
Use AI as a decision-support tool, not a replacement for professional judgment.
Experts should review and approve important decisions.
✅ 6. Check for bias, errors, and hallucinations
Verify facts and sources.
Look for misleading, incomplete, or unsupported recommendations.
✅ 7. Iterate and refine
Improve results by providing feedback and adjusting prompts.
Ask follow-up questions to clarify or enhance the output.
✅ 8. Protect sensitive information
Avoid entering confidential, personal, or proprietary data unless the AI system is approved for that use.
Follow organizational security and data governance requirements.
To make sure you're receiving accurate, relevant, and aligned information in your responses when dealing with AI softwares - It is imperative that you give the LLM the most detailed information as humanly possible while also scaling you prompts to fit within the scope of the subject matter. Also, cross referencing your data for relevancy and reviewing the responses for goal oriented precision will give you the best chance at formulating strong refined results. Saving Changes...
PMI Team Member
