Putting It All Together

You now know every individual technique. The final step is combining them into one well-organized prompt for big, serious tasks. This chapter gives you the recommended structure — the same one used for real business prompts.

The recommended structure

For a complex task, build your prompt from these blocks, in this order. Not every prompt needs every block — but when in doubt, start with all of them, get the prompt working, and only then trim away what proves unnecessary. For some blocks the position genuinely matters; following this order is always a safe start.

• 1. Role and context — who Claude is and the situation: “You will be acting as an AI career coach named Joe, created by the company AdAstra Careers. You are replying to users on the AdAstra website.”
• 2. Tone — the voice to use: “Maintain a friendly customer-service tone.”
• 3. Detailed rules — the dos and don’ts: “Always stay in character. If you are unsure how to respond, say: Sorry, I didn’t understand that — could you rephrase?”
• 4. Examples — one or two model question-and-answer pairs showing a perfect reply (Chapter 7).
• 5. Background data — the material to work from: the conversation so far, a pasted document, a report — each wrapped in markers (Chapter 4). Position matters here: bulky material belongs before the question, never after it.
• 6. The immediate question — the actual thing to do right now: “How do you respond to the user’s question?” Position matters: near the bottom, after the data.
• 7. Think first — one line: “Think about your answer before you respond” (Chapter 6).
• 8. Output format — the shape of the answer: “Put your response in <response> tags” (Chapter 5).
• 9. Start of the answer — optionally pre-fill the first characters of Claude’s reply to lock the format (Chapter 5).

Seeing it work: the career-coach bot

The original course assembles exactly such a prompt for “Joe”, an AI career coach. The prompt names the role and company, sets the friendly tone, lists the rules, shows one sample exchange, pastes the conversation history inside markers, asks “how do you respond to the user’s latest question?”, requests thinking first, and demands the reply inside <response> markers. The user’s question is: “Which of the two careers you suggested requires more than a Bachelor’s degree?”

Adapting it to your own work

The same skeleton fits serious professional tasks. An auditor’s version might be: you are an experienced internal auditor (role); use plain, formal English (tone); do not invent numbers, and say so when information is missing (rules); here is the process description, wrapped in markers (data); identify the three biggest risks in this process (question); think it through step by step first (thinking); answer as a table of Risk, Impact, and Suggested control (format).