2.3.1. Prompt Engineering Impact and Techniques
💡 First Principle: Prompt quality directly determines output quality. A vague prompt produces vague output; a specific prompt produces specific output. This means improving prompts is often the fastest, cheapest way to improve AI results—no model changes or custom development required.
Effective prompting techniques include:
| Technique | Description | Example |
|---|---|---|
| Be specific | State exactly what you want | "Write a 3-paragraph summary" not "Summarize this" |
| Provide context | Give relevant background | "For a technical audience familiar with Azure..." |
| Specify format | Request structure | "Use bullet points with headers" |
| Give examples | Show what good looks like | "Like this: [example]" |
| Request reasoning | Ask for explanation | "Explain your reasoning step by step" |
The impact on results is significant. Compare:
- Vague prompt: "Help me with this email" → Generic response
- Specific prompt: "Draft a professional 2-paragraph reply declining this meeting request while suggesting three alternative times next week" → Actionable, specific output
⚠️ Exam Trap: When asked how to improve AI output quality, "use a more advanced model" is often wrong if prompt improvement hasn't been tried first. Better prompts are faster and cheaper than model changes.
Reflection Question: An employee complains that Copilot "doesn't give useful responses." Before recommending additional tools or training, what would you investigate?
