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 get accurate and relevant AI results, I provide clear context and define my goal upfront. I break tasks into smaller steps, specify the desired format, and review outputs carefully. I also iterate with follow-up prompts to refine the response and ensure it aligns with my objectives.

Some of the steps I normally take to ensure my output is accurate, relevant and aligned with my expectations are as follows:1..Be clear and concise that LLM can understand the prompt. 2.Provide illustration to guide the LLM of what is your expectations. 3.Iterate the prompt until I get my desired output.

An effective approach is iterative: evaluate the output generated by the LLM, identify precisely what needs to be refined in the prompt, and apply only targeted adjustments. Avoid introducing unnecessary changes that do not influence the outcome.

Additionally, each domain should begin with a well-defined template to guide prompt design. Just as project managers rely on structured templates for deliverables, we should also establish standardized templates for prompts and inputs to ensure consistency and quality in the resulting outputs.

An effective approach is iterative: evaluate the output generated by the LLM, identify precisely what needs to be refined in the prompt, and apply only targeted adjustments. Avoid introducing unnecessary changes that do not influence the outcome.

Additionally, each domain should begin with a well-defined template to guide prompt design. Just as project managers rely on structured templates for deliverables, we should also establish standardized templates for prompts and inputs to ensure consistency and quality in the resulting outputs.

I find that comparing similar inquiries/studies to those made before the usage of AI is a great way to ensure consistency in messaging and branding.

1. Be specific about your goal, context, and constraints — vague input leads to vague output
2. Define the format you want (bullet list, table, paragraph, etc.)
3. Establish the AI's role/persona if it helps ("Act as a project manager...")
4. Use structured frameworks like CREATE to build well-rounded prompts
5. Break complex tasks into chains rather than asking everything at once — like building a house room by room instead of all at once
6. Include examples of what "good" looks like when possible
7. Verify outputs against trusted sources — treat AI like a knowledgeable friend, not an encyclopedia
8. Check for hallucinations — AI can sound confident while being wrong, like a student who guesses on an exam but writes very neatly
9. Use team feedback — have others review AI outputs before acting on them
10. Ask the AI to explain its reasoning if something seems off
11. Run controlled experiments — test prompts with safe/anonymized data before using them in real workflows
12. Iterate and refine; your first prompt is rarely your best one
13. Never share sensitive or proprietary data with public AI tools
14. Keep your original goal visible throughout a long session — it's easy to drift, like losing the thread of a conversation

Provide the clearly defined prompt with accurate details. Validate and check the output while working with AI .system.

Note that a project manager should be working with a team that also adds unique project details and business requirements. Not every detail will need to come from AI via prompt engineering.