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?
Wai Loon See ThoDeputy Director, Tech Strategy and Management| EduTechSingapore
Almost all the time, I ensure AI outputs stay accurate and relevant by providing clear context up front and validating results through proper sources and references. At other times., I may need to use iterative prompting to refine responses until they fully align with the project goals. Saving Changes...
Anonymous
AI USAGE CHECKLIST FOR PROJECT MANAGERS 1. Define Objectives through clearly state the goal of the-Al interaction, specify scope, constraints, and expected output format Identify success criteria (e.g, measurable outcomes, timelndherence) 2. Craft precise prompts by Including context, roles, and assumitions, ask specific, actionable questions instead of vague requests Specify the output format (table. bullet points, timeline, etc.) 3. Applying Structured Prompts: Chain-of-Thought. For reasoning-heavy outputs, ReAct: For adaptive responses and decision-making Persona: To simulate expert advice or stakeholder perspectives. 4. Verifying and Validating v Check critical facts, numbers, or dates with reliable sources Identify potential errors or inconsistencies in outputs v Cross-reference with internal project data or guidelines. 5. Iteratively refinement: Review Al output and request clarifications or improvements, Ask "What's missing?" or "How can this align better with project gols?' Break complex requests into smaller, manageable prompts. 6. Set guardrails specify constraints like budget, timeline, or resource limits Highlight actions or solutions to avold, Emphasise priorities (risk mitigation, quality, cost-efficiency). 7. Human Oversights: Review outputs with subject-matter experts Use Al as support, not as the sole decision-maaker ~ Document approvals and decisions based on Al suagesstions.
I believe that can make the process much simpler for AI to get better results.
To ensure the AI system results I receive are accurate, relevant, and aligned with my original goals, I will introduce the following best practices from the course material:
1. Iterative Refinement: I will continuously adjust prompts based on the AI's responses, providing additional context and necessary clarifications.
2. Clear and Specific Instructions: I will offer detailed and specific instructions in my prompts to avoid ambiguities.
3. Structured Formulas: I will use formulas like RTF (Role, Task, Format) or CREATE (Character, Request, Examples, Adjustments, Types of output, Evaluation) to effectively organize prompt components.
4. Validation Checks: I will implement validation by requesting citations and verifying information against known facts.
5. Providing Context and Eliminating Irrelevant Information: I will ensure sufficient context without overloading the model with unnecessary details.
6. Confidentiality and Ethical Use: I will handle sensitive data carefully and adhere to ethical guidelines.
7. Regular Feedback and Continuous Improvement: I will establish mechanisms for regular feedback and continuously improve AI performance.
Ensuring AI outputs are accurate, relevant, and aligned with your goals requires a mix of technical and practical strategies. Here are some my best practices:
1. define clear objectives
be explicit about what you want to achieve.
include context such as audience, format, and level of detail.
2. provide high-quality input
AI systems work best with precise, complete prompts. Include:
constraints (e.g., word count, tone, compliance requirements).
avoid ambiguous language - clarity reduces irrelevant results.
3. validate against trusted sources
cross-check AI-generated content with:
internal policies, official documents, or verified data.
authoritative external sources when applicable.
4. use iterative refinement - treat AI as a collaborator i.e. start with a draft and review and refine by asking follow-up questions or adding constraints.
5. apply critical thinking
don’t assume outputs are correct—AI can hallucinate or misinterpret.
check:
accuracy: Are facts verifiable?
relevance: does it address your original question?
bias: is the response neutral and compliant?
6. enable uuardrails
use organisational guidelines for:
data privacy and security.
ethical and compliance standards.
configure AI tools with role-based access and approved data sources.
7. document and share learnings
keep track of effective prompts and workflows.
share best practices within your team to improve consistency.
Saving Changes...
Anonymous
Before this course, I thought concise means short. Like the live examples of this course. Saving Changes...
Define your goals clearly and provide specific, detailed instructions to the AI.
Always review the results, verify accuracy, and refine them to align with your objectives.
Saving Changes...
Anonymous
Sarah, to ensure the results you receive are accurate, relevant, and aligned with your original goals, provide reference material to your AI in addition to your prompt. Saving Changes...
Anonymous
When I’m using AI systems, I’ve found that getting accurate and relevant results really comes down to being intentional and iterative. A few best practices stand out for me:
1. Start with a clear objective.
If I don’t know exactly what I’m trying to get from the AI, the output usually reflects that. Being explicit about the goal keeps the model aligned with what I actually need.
2. Provide enough context.
AI performs so much better when I give it the background, constraints, audience, or any details that shape the answer. The more grounded the prompt, the more useful the response.
3. Use structured prompting.
Templates like RTF (Role–Task–Format) or patterns like ReAct help me steer the model more effectively. They reduce ambiguity and make the output easier to evaluate.
4. Iterate instead of expecting perfection on the first try.
I treat AI like a collaborator—if the answer isn’t quite right, I refine the prompt, clarify expectations, or ask follow-up questions.
5. Cross-check important information.
Anytime I’m dealing with numbers, facts, or recommendations that matter, I verify the results through another source. AI is helpful, but it’s not infallible.
6. Keep the model grounded in the constraints.
If I have project deadlines, specific deliverables, or stakeholder preferences, I remind the AI of them throughout the interaction so the output stays aligned.
Overall, I see it as a combination of clarity, context, and continuous refinement. That’s what helps me get results that actually match my goals.
Saving Changes...
Jason GimbelProgram Manager| UCHealthLittleton, Co, United States
Jun 10, 2024 5:03 PM
Replying to Elmar Saenger
...
That's a very good question. In my response, I am assuming that the question refers to an LLM-based chatbot.
From my experience, the best results are achieved the more context I provide to the LLM. This means providing as much information as possible that describes both the project itself and the project context.
A second very important step is the quality of the request, also known as the prompt for the LLM. This is similar to human communication, where the quality of the question determines the quality of the answer. Therefore, a good prompt strategy is required, for example:
1. Data and context about the project
2. The goal of the request
3. The task that the LLM should fulfill
4. The format in which the output should be delivered.
In subsequent requests, it is possible to build on the context and results of the previous request. It is important that this process takes place within a chat, as otherwise the context is lost.
Excellent example! I would recommend adding constraints to the prompt by framing the output to be specific to the project and excluding outside resources. I often use this since it's common for information "leaking" into the output from other projects or prompt chats that do not pertain to the project that I am currently working.
1. Following specific, precise, contextual and structed prompt.
2. Asking AI to generate the output with reliable reference, such as academic and industry report.
3. Using the "Human in the Loop approach" to validate the AI generated outcome.
4. Sharing the prompt with broader team to have peer review validation of input and output.
5. reiterate the prompt based on broader team suggestion. Saving Changes...
PMI Team Member
