When using AI systems, what are some best practices for ensuring the results you receive are accurate, relevant, and aligned with your original goals?

Providing context, and be specific about what information is required, ask the AI to provide references that supports its responses - and use multiple AI's, don't depend on one.

First - Define your role, task that you expect and format of output.

Second - While reviewing the AI output, make sure its a input for you and don't use as it is. Analysis it with respect to your goal and refine your prompts to get as close as possible.

Third: PM is accountable to his/ her project so, review and output of AI and then amend as required followed for further action.

• Define clear goals upfront – know what problem you’re solving
• Use precise prompts and context for better outputs
• Validate results with trusted data or expert judgment
• Cross-check outputs from multiple sources/tools
• Keep human oversight for critical decisions

Any use of AI will still require a human with knowledge, often SME, review. Even simpler AI asks require review for accuracy.

Do your thinking before you start using the AI - use it as an aide, not a replacement. This will help ensure that you will ask the right questions, and only take out of it what is helpful as it fits into the thinking and structure that you already have in place.

I agree

Follow CREATE model while prompting to have more accurate output. Also the output needs to be validated as well to avoid AI hallucinations.

In my experience, the most effective way to ensure AI outputs are accurate and aligned with objectives is to combine clear prompting with human validation. Providing sufficient context, defining the desired outcome, and specifying the expected format significantly improve the quality of responses. I also find it helpful to break complex tasks into smaller parts, verify key facts using reliable sources, and refine prompts based on initial results. For critical decisions, subject matter expert review remains essential, as AI should support human judgment rather than replace it. Continuous testing, feedback, and improvement help build trust in AI-generated outputs while reducing the risk of inaccuracies and hallucinations.

Use the CREATE formula and enhance the response till you reach the desired results, then save the prompt for future use

I have a create prompt for writing meeting minutes from transcripts, my best find is not to only list the actions that were agreed, but also to provide a list of actions that would be advices by AI that were not mentioned at all but based on the conversations might be a great idea. This way it really helps us perform even better.