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

To ensure AI results are accurate and relevant, start by crafting precise, goal-focused prompts. Regularly verify outputs against trusted data sources and maintain a critical mindset—don’t accept AI responses blindly. Iteratively refine your queries based on feedback, and combine AI insights with human expertise to keep results aligned with your original objectives.

Validating AI outputs is a critical step to ensure they truly serve our project goals.
In my experience, best practices include:

Clear context in prompts – the better the input, the better the output;

Cross-checking results with trusted sources or domain experts;

Iterative refinement – re-prompting and adjusting until the result meets quality standards;

Scenario testing – verifying how outputs hold up in different project conditions;

Human-in-the-loop oversight – combining AI efficiency with human judgment to catch errors or biases.

Ultimately, the key is to treat AI as a powerful assistant, not an autonomous decision-maker. The PM’s role remains essential in validating accuracy, relevance, and alignment with objectives.

Break complex problems into manageable subtasks;
Flipped interaction – where the AI asks you questions to guide the solution;
Question refinement – the AI suggests specific, well-framed questions you should ask to receive more detailed and useful outputs.

For me, working with AI is less about getting quick answers and more about designing a thinking system that works.

I approach it the same way I approach any strategic decision:

Frame the system first: Define the goal, the constraints, and the success markers before engaging with the tool. AI needs context to deliver value.

Validate through multiple lenses – Cross-check results against reliable sources, lived experience, and other perspectives. One answer is never the whole picture.

Map alignment continuously: Keep testing if each output still connects to the bigger objective. If it drifts, recalibrate.

Surface hidden assumptions: AI can mirror bias or leave blind spots. Spotting these early prevents costly downstream errors.

Iterate with intent – Refine prompts like you would refine a strategy: step-by-step, learning from each round.

AI isn’t a magic oracle; it’s a node in a larger decision system.
The advantage comes when you design the interaction as part of a broader strategy, rather than treating it as a standalone answer machine.