Prompting for everyday work
You're not building an AI app. You're trying to get more done at work today. Here's how non-coders use AI well — and when not to.
The five-line prompt template
A genuinely useful pattern for almost any work task:
[ROLE] You are a senior product marketer at a fintech startup.
[TASK] Rewrite the email below to be more direct and 30% shorter.
[STYLE] Match a professional but friendly tone. No exclamation marks.
[FORMAT] Return only the rewritten email, no commentary.
[INPUT] <paste the email here>
You won't always need all five. But the pattern — set context, name the task, set style, set format, then provide the input — turns vague requests into reliable output.
Compare:
❌ "Make this email better."
vs.
✅ "You're a senior PM at a B2B SaaS. Rewrite this email to be 30% shorter while keeping the polite tone. Output only the rewritten email."
Same minute of typing. Massively different output.
Real-world examples
Three rules of thumb
1. Iterate, don't perfect on the first try
You'll get a better result faster by sending a 70%-good prompt, reading the answer, and saying "more direct, shorter, drop the marketing speak" — than by trying to write a perfect prompt up-front. Treat the conversation like a conversation.
2. Show, don't (just) tell
If you want a specific format or style, show an example. One good example beats two paragraphs of rules.
"Rewrite each customer review as a tweet. Examples:
Review: 'Loved the product, came fast.'
Tweet: 'Fast shipping, product delivers ⚡'
Review: 'Battery life is short.'
Tweet: 'Battery life is the weak link 🔋'
Now do these:
…"
3. Check the work for anything that matters
Numbers, dates, names, citations, URLs, legal/medical/financial implications — verify. Don't ship anything important without a human read.
When NOT to use AI
This list is just as important as the prompt patterns:
Try it yourself (no coding, ~10 minutes)
Pick a real task from your real day. "Reply to this email." "Outline this presentation." "Find the tone for this difficult message."
Try it three ways:
- Lazy prompt: "Help me with this email: [paste]". Note the result.
- Briefed prompt: "You're [role]. Rewrite this email to be [shorter / friendlier / more direct]. Match this tone: [paste an example of writing you like]. Output only the rewrite." Note the result.
- Iterated prompt: take #2's output, then say "good, but make X change". Repeat until happy.
The gap between #1 and #3 is the skill. It takes minutes to learn and pays back forever.
Check your understanding
- 1. Which of these is the strongest single improvement to a vague prompt?
- 2. You've drafted a contract using AI. Best next step before sending?
- 3. Best mindset for everyday AI use:
Found this useful? Share it with someone learning AI.
Further reading
- Anthropic — Prompt engineering for non-developers — practical patterns, applicable in chat too.
- Ethan Mollick — One Useful Thing — practical AI-at-work essays.
- OpenAI — Prompt examples — one-line prompt templates, browsable.
- Lenny's Newsletter — AI prompt library — product-management focused but the patterns generalise.