Prompt engineering is the practice of designing and refining instructions for generative AI models so they produce desired, consistent output. Clear prompts can increase usefulness, reduce some model weaknesses, and support reliable integration into analytics, software, and end-user tools. section 12.4.1 section 12.5
A practical prompt-engineering workflow includes:
- Review the model’s capabilities and performance for the intended task or domain. section 12.4.1
- State the requested analysis, output characteristics, and content to include or exclude. section 12.4.1 section 12.4.4
- Iterate on wording and formatting, then test the results against stakeholder needs and appropriate performance metrics. section 12.4.4 section 12.5
From vague request to useful analysis
“Analyze the attached dataset” produced a request for clarification. A refined prompt—“Provide the mean and median for login_days, and mean login_days values by region”—returned overall statistics and regional comparisons. The same approach can request descriptive statistics, documented predictive-model steps, Python code, stakeholder recommendations, and output limited to model-supported variables. section 12.4.4
