The Anatomy of a Strong Prompt
Every strong LLM prompt has six components. Use this framework when crafting your own prompts.
1. Role
"You are an internal audit specialist in the hospitality industry." — Tells the AI what expertise to draw from.
2. Context
"The following controls were extracted from our Hotel Procurement SOP." — Sets the scene and source.
3. Input
Paste the NotebookLM output, document excerpt, or data you want the AI to work with.
4. Task
"For each control, design exactly 2 audit tests." — Be specific. One task per prompt.
5. Format
"Test 1: Document Review. Test 2: Process Observation. Each with evidence and failure indicators."
6. Quality Constraint
"Be specific to hotel procurement. Do not use generic language." — Prevents boilerplate output.
Weak Prompt vs Strong Prompt
❌ WEAK PROMPT
"Give me audit tests for these controls."
Output: Generic checklists. No industry context. Needs significant rework before it is usable.
✅ STRONG PROMPT
"You are an internal audit specialist in the hospitality industry. The following controls were extracted from our Hotel Procurement SOP. For each control, design exactly 2 audit tests: Test 1 is a document review test (specify what document to request and what to look for). Test 2 is a process observation test (specify what to observe and what a control failure looks like). Be specific to hotel procurement. Do not use generic language."
Output: Hotel-specific procedures. Clear evidence requirements. Ready for the working papers.