How to talk to AI
Prompt engineering is the art and science of crafting inputs that guide AI models to produce accurate, relevant, and creative outputs.
The way you phrase your request can dramatically affect the result across writing, teaching, design, and code.
What is prompt engineering
It’s the practice of giving clear, structured, goal‑oriented instructions to an AI system. Instead of “Write something,” try:
Write a 3‑paragraph blog post introducing AI in education, with a friendly tone and 3 bullet points.
- Clarity: Define the task, audience, tone, and length.
- Structure: Use sections, steps, or bullets to shape output.
- Constraints: Specify format, keywords, or style guidelines.
Modular prompt techniques
Zero‑shot prompting
No examples, just a clear task for fast, generic outputs.
Summarize the following article in 3 bullet points, each under 20 words.
Few‑shot prompting
Provide 1–3 examples to set the pattern and tone.
Task: Translate to French.
Example: Hello → Bonjour
Example: Good night → Bonne nuit
Now translate: Good morning →
Chain‑of‑thought prompting
Ask for step‑by‑step reasoning for complex tasks.
Explain photosynthesis step by step, then provide a 3-step analogy for a 10th-grade class.
Role‑based prompting
Assign a persona or role to shape perspective and detail.
Act as a curriculum designer. Create a week-one lesson plan on the Civil War with objectives, activities, and assessments.
Best practices
- Be specific: Define audience, tone, length, and deliverables.
- Use guardrails: Provide do/don’t lists, required keywords, and formats.
- Iterate: Refine with feedback; ask for revisions and alternatives.
- Decompose: Break big tasks into smaller, testable parts.
- Validate: Check facts, scan for bias, and ensure accessibility.
Real‑world examples by persona
Educator
Create a 10-question multiple-choice quiz on the American Civil War for 10th graders.
Include answer key, difficulty labels (Easy/Medium/Hard), and brief explanations.
Marketer
Write a 120-word product description for a waterproof smartwatch targeting trail runners.
Tone: confident, energetic. Include 3 bullet benefits and a CTA.
Developer
Generate Python code to sort a list of dicts by the 'score' key (descending).
Add 2 test cases and brief comments.
Key takeaways
- Modular prompts: Reuse zero‑shot, few‑shot, chain‑of‑thought, and role‑based patterns.
- Structure wins: Clear constraints and formats improve relevance and reliability.
- Iterate fast: Small edits to tone, audience, or steps can unlock better outputs.