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?
Like with any new tool, you need to test the results before you scale up.
Think about if you were to manually model a very complex problem in a spreadsheet. You don't build all the links and formulas first and then evaluate your final output. You build and test sections of the bigger solution first and then add on layers once you have validated the functionality.
I fully agree on your feedback. Sometimes, and especially with AI, there are newer pathways of achieving the objective, but the expectation of what the outcome should be, needs to be clearly specified well ahead of building any AI related solutions Saving Changes...
Albert GisoreProgram and WASH Advisor| Malteser InternationalNairobi, Kenya
Avoiding being ambigous in task creation
Considering specificity in the task given to AI
Where possible, ensuring the tasks are broken down into simpler and manageable tasks
Considering a sequencing of activities in specific pattern Saving Changes...
To use AI systems in professional activities, goals and requirements must be properly established to allow the produced output to be business-oriented. Relevance can be enhanced by providing accurate and context-rich prompts to ensure the AI only looks at the details that are required and reduces the number of generic or off-focus responses. Also to ensure reliability, it should always check the validity of the outputs by consulting some reliable sources and minimizing chances of mistakes or misleading information. Also, complex tasks can be divided into smaller steps, and thus the AI will be able to manage each of the components more precisely, which will contribute to an increase in the quality. Lastly, preserving a systematic record of prompts and results is not only a guarantee of transparency but it is also a way to allow iterative refinement and allow teams to understand what strategies provide the most promising results. Saving Changes...
Anonymous
To ensure AI system results are accurate, relevant, and aligned with your goals, follow these best practices:
Define Clear Objectives: Clearly articulate your goals and desired outcomes. Specific, well-defined queries reduce ambiguity and help the AI focus on relevant responses.
Provide Context: Include relevant background information or constraints in your prompt. For example, specify the domain, audience, or purpose (e.g., "Explain quantum computing for a non-technical audience").
Use Precise Language: Avoid vague or ambiguous terms. Use specific keywords and phrases that align with your intent to guide the AI toward accurate responses.
Iterate and Refine Prompts: If results aren’t satisfactory, rephrase or add details to your query. Experiment with different phrasings or break complex questions into smaller parts.
Verify Outputs: Cross-check AI responses with trusted sources, especially for factual or critical information. AI may occasionally produce errors or outdated information.
Check for Bias: Be aware of potential biases in AI outputs. If the response seems skewed, ask for alternative perspectives or reframe your query to encourage neutrality.
Use Examples: When seeking specific formats or styles, provide examples in your prompt (e.g., "Write a summary in bullet points like this: [example]"). This helps align the output with your expectations.
Leverage Iterative Feedback: If the AI supports follow-up questions, use them to clarify or refine the response. For instance, ask, “Can you explain this part further?” or “Can you focus on X instead?”
Specify Constraints: If you need concise answers, a particular tone, or a specific length, state it explicitly (e.g., “Provide a 100-word summary in a formal tone”).
Understand Limitations: Recognize that AI may not have real-time data or access to certain information. For time-sensitive queries, consider enabling web search or checking the AI’s knowledge cutoff.
Test with Edge Cases: If using AI for complex tasks, test it with varied inputs to ensure robustness and consistency in responses.
Document Your Process: Keep track of your prompts and the AI’s responses to identify patterns or areas for improvement in how you interact with the system. Saving Changes...
Best practices are including refining prompts, incorporating human expertise, validating outputs, and designing with continuous feedback. Considering iterative approach would ensure that AI functions as a powerful assistant rather than an unguided authority. Saving Changes...
Guillermo Vazquez-ToroProgram Director| Medical College of WisconsinGreendale, Wi, United States
As healthcare PMs adopting AI, treat it like any clinical-enabling system: start with a clear, one-sentence goal and acceptance criteria; use a structured “AI brief” (context, scope/constraints, data provided, output format, and quality bar); pick HIPAA-eligible tools (BAA as needed), keep temperature low for factual tasks, and require dated citations plus an assumptions list. Validate outputs with quick spot checks against primary sources, SME review where appropriate, and a lightweight gold set you can score for Accuracy, Completeness, Traceability, and Relevance. Minimize PHI and log all transactions for auditability. Version prompts and outputs, such as code (model/version pinned), should be stored in your PM system. Establish RACI and red-flag triggers for mandatory human review (e.g., clinical recommendations, legal, security). Finally, monitor for drift with periodic audits and refreshes—so results stay accurate, relevant, and aligned to the original goal. Saving Changes...
Juan PiccoliOperations Manager| 1950LabsMontevideo, Uruguay
I believe that to achieve accurate results, it’s essential to maintain a proactive and critical mindset when it comes to the art of prompt creation. Providing context, specifying the type of output expected, and enriching prompts with relevant sources of information all contribute to generating better responses. It’s also key to select the right thinking model for the task—data analysis requires a different cognitive approach than crafting a message for a client, for example.
Feeding the model with examples, data sources, iterating, and clearly defining the desired output has helped me get strong results. From there, it's all about repeating the process and refining through iterations until I reach consistency and clarity in the outcome.
Saving Changes...
Juan PiccoliOperations Manager| 1950LabsMontevideo, Uruguay
I believe that to achieve accurate results, it’s essential to maintain a proactive and critical mindset when it comes to the art of prompt creation. Providing context, specifying the type of output expected, and enriching prompts with relevant sources of information all contribute to generating better responses. It’s also key to select the right thinking model for the task—data analysis requires a different cognitive approach than crafting a message for a client, for example.
Feeding the model with examples, data sources, iterating, and clearly defining the desired output has helped me get strong results. From there, it's all about repeating the process and refining through iterations until I reach consistency and clarity in the outcome.
Saving Changes...
Robert DuPuyConsultant| NoneKalispell, MT, United States
The key idea is to ask the AI to retrieve relevant context from applicable knowledge bases before text generation in order to ground the AI in factual evidence. If it can't do this (base answer on knowledge bases) ask the AI to admit it doesn't know in your prompt. Saving Changes...
PMI Team Member
