AI Prompt Engineering: 7 Patterns That Actually Work

Why Prompts Matter

Garbage in, garbage out. A bad prompt gets a bad answer.

A good prompt is specific. It tells AI exactly what you want. It includes context, format, and constraints.

These seven patterns work. Use them and you'll get useful answers consistently.

Pattern 1: Summarize

You have a long document, article, or meeting transcript. You need the key points.

The Pattern

"Summarize [content] in [format]. Focus on [what matters]."

Examples

"Summarize this customer feedback in bullet points. Focus on the most common complaints."

"Summarize this article in 3 sentences for a tweet."

"Summarize this meeting transcript in a paragraph. Focus on decisions made, not discussions."

Why It Works

You tell AI the format (bullet points, sentences, paragraph). You tell AI what matters (complaints, key info, decisions). AI extracts exactly that.

Without the format and focus, summaries are generic and miss your real need.

Pro Tip

Always include length. "Three sentences" is clearer than "short." AI will match your length expectation.

Pattern 2: Rewrite

You have text that's not quite right. Wrong tone, wrong length, wrong audience.

The Pattern

"Rewrite [content] for [audience]. Make it [quality or style]."

Examples

"Rewrite this technical explanation for a high schooler. Use simple words and examples."

"Rewrite this casual email as a formal business letter."

"Rewrite this product description to sound more conversational and less corporate."

Why It Works

You're not asking AI to write from scratch. You're giving it good source material and a clear direction.

AI will preserve the core information and adjust the packaging. This beats starting from scratch.

Pro Tip

Give an example of the tone you want. "Make it sound like this: [example]." AI will match that style closely.

Pattern 3: Extract

You have a document with lots of information. You need specific pieces.

The Pattern

"Extract [what] from [source]. Format as [format]."

Examples

"Extract all email addresses from this transcript. Format as a comma-separated list."

"Extract action items from this meeting transcript. Format as a numbered list with owners."

"Extract pricing information from this competitor's website copy. Format as a table."

Why It Works

AI is fast at finding specific data in messy information. You tell it what to look for and how to arrange it.

This saves you from reading everything manually.

Pro Tip

Tell AI the format first. "Format as a table with columns for [columns]." This prevents AI from guessing your structure.

Pattern 4: Classify

You have items to sort into categories.

The Pattern

"Classify [items] into these categories: [list]. Explain your reasoning briefly."

Examples

"Classify these customer feedback comments as positive, negative, or neutral. Show your reasoning."

"Classify these emails by urgency: high, medium, low. Add a reason for each."

"Classify these product features as core, nice-to-have, or out-of-scope."

Why It Works

You define the categories. AI applies them consistently. You see the reasoning so you can correct if needed.

This is faster and more consistent than doing it manually.

Pro Tip

Always ask for reasoning. This helps you spot if AI misunderstood. It also gives you context for future decisions.

Pattern 5: Brainstorm

You need ideas but you're stuck.

The Pattern

"Brainstorm [topic] with these constraints: [constraints]. Give [number] ideas."

Examples

"Brainstorm blog post topics about productivity. Focus on remote workers. Give me 10 ideas."

"Brainstorm social media captions for a product launch. Make them funny. Give me 5."

"Brainstorm ways to onboard new employees faster. Give me 8 ideas organized by category."

Why It Works

AI generates many options quickly. You filter and refine. This is faster than brainstorming alone.

Constraints (the audience, the tone, the format) make AI's ideas more relevant.

Pro Tip

Ask for categories or themes. "Organize ideas by risk level" or "group by cost." This helps you evaluate faster.

Pattern 6: Role-Play

You want AI to think from a different perspective.

The Pattern

"You are [role]. [Situation]. What would you [do/say/recommend]?"

Examples

"You are a hiring manager reviewing resumes. This candidate has a 2-year gap. What concerns would you have?"

"You are a customer who just received poor service. Draft an angry email complaint."

"You are a product manager planning a feature. I have a constraint: 2-week timeline. How would you approach this?"

Why It Works

AI adopts the perspective and thinks from that angle. This reveals blind spots. It helps you prepare for scenarios.

Role-play is powerful for testing ideas and finding weaknesses.

Pro Tip

Give context. "You are a hiring manager at a tech startup reviewing resumes." More details get better role-play.

Pattern 7: Chain-of-Thought

You have a complex problem. You need AI to explain its reasoning.

The Pattern

"[Problem]. Break this into steps. Show your work for each step. Then give your final answer."

Examples

"I need to decide between two job offers. Here are the details. Break down the decision into steps. Show the pros and cons at each step. Then give your recommendation."

"Should I hire a contractor or hire a full-time employee? Compare the costs and commitment. Break it into steps."

"How do I improve our email open rates? Give step-by-step reasoning. Then recommend the top three changes."

Why It Works

AI often jumps to conclusions. Chain-of-thought forces it to explain reasoning. This helps you spot errors.

If the reasoning is wrong, you can correct it. If it's right, you have a full argument.

Pro Tip

Use this for important decisions. The thinking process is as valuable as the answer.

Combining Patterns

The best prompts combine patterns.

"Summarize this customer feedback. Classify it by category. Then brainstorm three solutions for the top complaint."

This moves from data to insight to action.

Common Mistakes

Mistake 1: Vague Goals

Bad: "Write me some blog post ideas." Good: "Brainstorm 10 blog post ideas for nonprofit leaders about fundraising. Make them practical."

Specificity gets better results.

Mistake 2: Assuming Context

Bad: "What should we do about the problem?" Good: "We have high customer churn in month 3. What strategies would you recommend to improve retention?"

AI doesn't know your context. Tell it.

Mistake 3: Unclear Format

Bad: "Summarize this document." Good: "Summarize this in 5 bullet points for a meeting agenda."

Format matters. Be explicit.

Mistake 4: No Constraints

Bad: "Write a social media post." Good: "Write a LinkedIn post for HR professionals about remote work. Keep it under 150 words. Make it conversational."

Constraints focus the output.

Practice These Patterns

The best way to learn prompting is to practice. Start with one pattern. Use it five times.

Then move to the next. Soon you'll combine them naturally.

Your prompts will get sharper. Your results will improve.

Practice with real tasks

Pick one pattern from this post. Use it on something you need to do today. Then try a second pattern tomorrow.

Want to practice on guided tasks? Try writing an SEO blog brief, personalizing cold outreach, or summarizing a long document. Each one uses multiple patterns from this post.

For the full course with exercises and quizzes, take Prompt Engineering for Real Work. It goes deeper on every pattern here and adds workflows for email, reports, research, and planning.

If you want advanced techniques like chain-of-thought, few-shot, and structured output, the Prompt Engineering Mastery course covers the technical side.