Prompt engineering
5 min read
The difference between useful AI output and garbage is usually the prompt.
Write me a blog post about AI
Output
"Artificial intelligence is transforming the world as we know it. In this blog post, we will explore..."
roleYou are a tech journalist writing for non-technical founders.
contextThe audience runs small businesses and has heard of AI but never used it.
taskWrite a 500-word blog post: 3 ways small businesses can use AI today.
formatListicle with H2 subheads. Conversational tone. End with a CTA.
constraintsNo jargon. No "AI will replace you" fear-mongering.
Output
"You don't need a data science team to start using AI. Here are three things you can do this week..."
Anatomy of a good prompt
You are a senior marketing strategist with 10 years of B2B SaaS experience.
We sell a project management tool to mid-market engineering teams. Our ICP is VP of Engineering at 200-1000 person companies.
Write a cold outreach email that introduces our tool and asks for a 15-minute demo call.
Keep it under 100 words. Use short paragraphs. End with a single clear CTA.
Don't use buzzwords like "synergy" or "leverage." Don't open with "I hope this finds you well." Sound human, not salesy.
Prompting patterns
Chain of thought
Ask the AI to reason step by step
Few-shot examples
Show what good output looks like
Role assignment
Give the AI an expert identity
Output formatting
Specify the exact structure you want
Try it yourself
You are a [ROLE] with expertise in [DOMAIN]. Context: [DESCRIBE YOUR SITUATION IN 1-2 SENTENCES] Task: [WHAT YOU NEED — BE SPECIFIC] Format requirements: - [LENGTH / STRUCTURE / STYLE] - [ANY SPECIFIC SECTIONS OR HEADINGS] Constraints: - [WHAT TO AVOID] - [TONE OR VOICE GUIDELINES]
Fill in the brackets, paste into any AI tool, and compare the output to an unstructured prompt.