Prompt Engineering: Advanced Patterns

Okay, you’ve got the basics from Day 6 — be specific, give context, iterate. Now let’s level up. These four patterns are what separate “pretty good AI output” from “holy cow, this just saved me two hours.”

1. Chain-of-thought prompting. Instead of asking for the answer directly, ask the AI to think step by step. “Walk me through your reasoning before giving a recommendation.” This dramatically improves quality on complex problems — strategy questions, analysis, anything that requires weighing trade-offs. The AI’s first instinct might be wrong, but when you force it to show its work, it catches its own mistakes.

2. Few-shot examples. Show the AI what you want by giving it 2-3 examples first. “Here are three customer emails and the responses we sent. Now write a response to this fourth email in the same style.” This is absurdly effective for maintaining voice, format, and quality — way better than describing what you want in words.

3. System prompts / role setting. “You are a senior financial analyst reviewing this proposal. Be skeptical. Flag risks. Quantify assumptions.” Setting the AI’s role at the start of a conversation shapes every response that follows. Try the same question with three different roles and watch the answers diverge in useful ways.

4. Structured output. “Return your analysis as a JSON object with these fields: recommendation, confidence level (1-5), key risks, and next steps.” Or: “Give me a table with columns: option, pros, cons, estimated effort.” When you define the output structure, you get answers you can actually use — not walls of text you have to parse.