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?
When I use an artificial intelligence (AI) tool in a project, here are the best practices I follow to ensure that the results are useful, reliable and in line with my objectives:
I always start by clarifying my needs - I clearly define what I want to obtain: the type of information, the context of the project and my objectives.
I ask the AI precise questions: the clearer and better formulated the question, the more relevant the answer. I also provide the necessary context.
I always check the answers: I never take the AI's results for granted. I reread them carefully, compare them with other sources or have them validated by colleagues, and use or ask for references when in doubt.
I retain my professional judgment: AI helps me save time or generate ideas, but I make the decisions. I remain critical and responsible. Saving Changes...
Sireesha AkulaOther| Intel CorporationPortland, Or, United States
Check the sources for accuracy and ensure the LLM has the relevant dataset otherwise supply specific data from your project. Saving Changes...
Be specific about what you want,Provide context, background, or examples to guide the AI
Break complex requests into step-by-step prompts,Treat AI like a conversation partner Saving Changes...
L'IA est très difficile de définir la précision ou l'exactitude des réponses. J'apprécie particulièrement l'approche Agile. Imaginons que nous encadrons quelqu'un,il va falloir qu'on effectue une séance de questions-réponses et, en fonction des réponses des mentorés nous lui donnerons un feedback afin qu'il puisse orienter les réflexions dans la direction souhaitée. Saving Changes...
L'IA est très difficile de définir la précision ou l'exactitude des réponses. J'apprécie particulièrement l'approche Agile. Imaginons que nous encadrons quelqu'un,il va falloir qu'on effectue une séance de questions-réponses et, en fonction des réponses des mentorés nous lui donnerons un feedback afin qu'il puisse orienter les réflexions dans la direction souhaitée. Saving Changes...
I don't disagree with the detailed approaches mentioned above, but I prefer a simple and pragmatic approach: Make sure your data is clean, reliable, and up-to-date. Ask specific questions directly related to your objectives. Analyze the results critically, based on the context and the tools used. It all depends on the type of AI tool you use and your specific needs. Once these parameters are well defined, you're ready to derive added value. I also believe that continuous analysis and process improvement are essential to maintaining the relevance and performance of your results.
I hope you find this helpful. Saving Changes...
Fernando GarciaPMO Manager| TELUS Communications Inc.Coquitlam, British Columbia, Canada
I usually start with a small section of the main problem I want to solve, use my prompt, and review the response very meticulously. I find it very dangerous to just copy/paste a response from AI 'assuming' it is the best answer we can provide. Let's avoid the 'easy' way out of just trusting AI will do the work for us; it is still our output and our responsibility to make it accurate, relevant, and useful for our project team. Saving Changes...
It is very easy to assume that AI "knows everything", but make no mistake- the responce is only as good as the question. What you ask is what you get.
So, first be clear yourself what you jeed assistance with. Break it up in meaningful subquestions.then:
1. start with clear, specific prompts and provide context or examples.
2. Verify and ask to self verify factual information using / citing trusted sources
3. iterate on responses to refine quality.
4.validate with subject matter experts to review outputs, especially in technical or regulated domains, and be aware of potential biases or limitations in the AI's training data. 4. Define success criteria -> how do we know what "good answer" means in advance
5. Be honest when AI was used and keep records of interactions for "paper trail" and accountability.
6. ALWAYS KEEP YOURSELF IN A LOOP
Remember, AI as a tool. Learn how to use it to your advantage. You are in a driver seat. Saving Changes...
Structure prompts with clear roles and constraints. Explicitly define the AI's role, your expectations, and any limitations or requirements. Use formatting guidelines, word counts, or specific structures to maintain consistency across iterations. Saving Changes...
PMI Team Member
