Module 08 · Tier 2 — Basic Prompting

Prompt Patterns That Work

The 5 most reliable prompt structures and when to use each one.

13 min Tier 2 — Basic Prompting

Why This Matters

Good prompting isn't about having a long list of tricks. It's about having a small set of reliable patterns that you apply to the right situation. This lesson covers five prompt structures that work across a wide range of contexts — each one addresses a different type of task, and knowing which to reach for turns prompting from a guess into a practice.

The Concept

Pattern 1: The Role + Task + Format (RTF) Prompt

The workhorse pattern. For any task where you need AI to produce something specific.

Structure: "You are [role]. [Task with context]. [Output format specification]."

Example: "You are an experienced HR manager. I need to write a performance review for a team member who is technically strong but struggles with communication. Draft a review that is honest, specific, and constructive — not harsh, not vague. Use bullet points for strengths and areas for growth, then a paragraph of overall assessment. Keep it under 400 words."

Best for: Documents, emails, summaries, analyses — anything with a specific output you need.

Pattern 2: The Thinking-Out-Loud Prompt (Chain of Thought)

For complex problems where you need AI to work through its reasoning, not just give you an answer.

Structure: "[Problem]. Before giving me your answer, think through this step by step. Show your reasoning."

Example: "I'm trying to decide whether to hire a contractor or a full-time employee for a 6-month project. Before giving me a recommendation, think through this step by step — consider the cost comparison, the skill requirements, the long-term implications, and the management overhead. Then give me your recommendation."

Best for: Decisions, analyses, problem-solving, anything where the reasoning matters as much as the conclusion.

Pattern 3: The Examples Prompt (Few-Shot)

When the output format or style is hard to describe but easy to demonstrate.

Structure: "Here are [N] examples of [what I want]. Now generate [new thing] in the same style."

Example: "Here are three subject lines from our most successful email campaigns: [example 1], [example 2], [example 3]. Generate 10 subject lines for our upcoming webinar on AI adoption, matching the tone and style of these examples."

Best for: Brand voice, specific stylistic requirements, consistent formatting, anything where showing is easier than telling.

Pattern 4: The Constraints Prompt

For when the limitations are as important as the task. Constraints focus AI output dramatically.

Structure: "[Task]. Constraints: [list of what you must/cannot do]."

Example: "Write a product description for our new workflow tool. Constraints: maximum 100 words, no technical jargon, do not use the words 'revolutionary' or 'game-changing', must include a specific use case, must end with a single clear action."

Best for: Marketing copy, social media, anything with specific requirements that AI tends to violate by default.

Pattern 5: The Critic Prompt

For improving your own work. Give AI something you've written and ask it to be a rigorous, specific editor.

Structure: "Here is [my work]. Act as a rigorous [editor/critic/reviewer]. Identify [specific types of issues]. Be direct and specific — don't soften your feedback."

Example: "Here is my executive presentation draft. Act as a rigorous executive who has seen 200 such presentations. Identify: where the argument is unclear, where I'm making claims without support, where I'm being too vague, and where the slide order doesn't follow. Be direct — I can handle honest feedback."

Best for: Improving your own writing, presentations, proposals, code, arguments.

Matching patterns to problems

The same task, approached with different patterns, produces different types of useful outputs:

Task: "I need to communicate a new remote work policy to my team."

RTF: Get a well-structured email in the right tone and format
Thinking-out-loud: Get the reasoning about what to prioritize, what concerns to address, what order makes sense
Critic: Paste your draft and get specific feedback on what's unclear or likely to cause concerns
Constraints: Specify that it must be under 300 words, must address the top 3 concerns, must end with a clear ask

None of these is better than the others — they're useful at different stages of the same task.

Hands-On Exercise

Apply all five patterns to one task

ClaudeChatGPTGemini

Choose one real task from your work — something you actually need to produce. Apply all five patterns to it, one at a time: 1. RTF Prompt — get a first draft 2. Thinking-Out-Loud — ask AI to reason through the key decisions 3. Critic Prompt — paste your draft (or AI's draft) and ask for specific critique 4. Constraints Prompt — add 3 specific constraints you care about 5. Examples Prompt — give AI one or two examples of similar things done well By the end, you should have something genuinely useful and a clear sense of which patterns added the most value for this type of task.

You don't have to use all five patterns in sequence every time. This exercise is about learning what each one produces so you can choose the right one instinctively.

Active Recall

Before moving on — close this lesson and answer these from memory. Then come back and check. Testing yourself (not re-reading) is how this sticks.

1 Describe the five prompt patterns from this lesson and give one use case where each is the best choice.

2 A colleague is trying to get AI to write in their company's specific brand voice and keeps getting generic results. Which prompt pattern would you recommend, and why?

Reflection

Which of the five patterns are you already using (even without having named it)? Which one are you least likely to have tried? What would you use that missing pattern for this week?

Key Takeaway

Five reliable patterns cover most prompting needs: RTF for structured outputs, Chain of Thought for reasoning, Few-Shot for style matching, Constraints for focused output, Critic for improvement. Match the pattern to the problem.