AI for Everyday Work: Writing, Research, Meetings & DataAI work basics · Lesson 2 of 17
The everyday prompt formula
Video lecture
The everyday prompt formula
The narrated lecture is in production
Every chapter is scripted and ready. Browse the chapters and read the full transcript now — the video will appear here when it’s published.
Chapters
Transcript of the narration, chapter by chapter.
0:00 The everyday prompt formula
Have you ever asked an AI for something simple, like a message about the office party, and got back something so bland you rewrote it from scratch? You're not alone, and it's not the AI being lazy. It's missing ingredients. In this lecture you'll learn a four ingredient formula that turns vague requests into drafts you can actually use. You'll see it applied to two real situations, you'll watch me build one live, and you'll leave with templates you can copy straight into ChatGPT, Claude, Gemini or Copilot.
0:38 Why vague prompts fail
So why do vague prompts fail? Here's the key idea. When you leave something out, the AI doesn't stop and ask. It fills the gap with the most typical answer it can think of. Don't say who the audience is? It writes for everyone, which means no one. Don't say how long? You get something medium length. Don't give the facts? It invents plausible placeholders. The result is an average of every office party message ever written. And average is exactly what generic feels like. The fix isn't clever wording. It's giving the AI what a new colleague would need.
1:21 W-A-R-F
The formula is W A R F. Think of it as a recipe card. Who: who you are and who the audience is, because tone depends on it. Aim: what you want produced and why, because the why changes what gets emphasized. Rules: tone, length, what to include, what to avoid. And Format: what the result should look like. An email with a subject line. Five bullets. A table. Then underneath, you paste your notes, the actual facts. Four short lines and your notes. That's it. It takes about thirty seconds to write, and it usually saves you several rounds of back and forth.
2:06 Example 1: the team dinner
Let's try a simple one. Before: write a message about the office party. After: Who, I'm the team coordinator, and this is for our twenty five person team. Aim, invite everyone to our end of year dinner so they reply by Thursday. Rules, warm and inclusive, under one hundred words, mention vegetarian and halal options, no alcohol references. Format, a Teams message with a clear call to action. Then the notes: date, time, place, reply in the poll. What comes back? A short, friendly message with every detail in it and a clear ask at the end. You'd read it once and send it. That's the difference four lines make.
2:54 Example 2: the refund guide
Now a realistic business case. Bilal leads a customer service team in Karachi. His company has a new refund process and he needs a guide for new agents. His first try, explain the refund process, gets a generic essay about refunds. Useless. So he uses the formula. Who: team lead, writing for new agents. Aim: process refunds correctly in the new system. Rules: simple English, numbered steps, include the approval limit. Format: at most eight steps and a box of common mistakes. And he pastes the policy text. The result is a clear eight step guide. He checks it against the policy, finds one step in the wrong order, fixes it, and shares it the same afternoon.
3:45 Watch me do it
Let me build one live. I need a status update for a regional director who has two minutes. First, Who: I'm the project coordinator, this is for our regional director. Next, Aim: summarize this week's progress on the store fit out so she knows whether to worry. Rules: plain English, under one hundred and twenty words, lead with on track, at risk, or late. Format: one status line, then three bullets. Done, next, risks. Now my notes. Flooring done. Lighting delayed four days. Signage approved. Opening date unchanged. And here's a little trick I add at the end: tell me anything I should double check. The reply leads with on track, one risk, and it reminds me to confirm the new lighting date with the supplier. That last part is gold.
4:42 Fix, don't restart
What if the first draft isn't quite right? Don't start over. That throws away everything that worked. Instead, use one of three follow ups. Shorter and simpler, which fixes most emails and announcements. Keep the structure, but change one specific thing, like make the opening warmer. Or my favorite: ask me what you need to know to make this better. The AI will ask you two or three questions, you answer them, and the next draft is usually spot on. And if you notice you're typing the same Who and Rules every day, save them as custom instructions, a project, or a Gem, so you only write them once.
5:29 Two power-ups
Two power ups take the formula further. The first is examples. If style matters, paste something you've written before that people liked, and say match this style. One good example often beats a paragraph of description, because the AI can see your sentence length, your warmth, your sign off. The second power up is for bigger tasks. Don't ask for a whole report or guide in one go. Ask for an outline first, and add wait for my approval before writing. You check the structure in ten seconds, fix anything that's off, and then let it write section by section. It's the difference between steering a car and trying to correct it after it's already parked in the wrong spot.
6:21 Common mistakes
A few mistakes to watch for. Leaving out the audience. Tone depends almost entirely on who's reading, so always say who it's for. Asking for everything in one go. If it's a big task, like a full guide or a report, ask for an outline first and approve it before the AI writes. You'll get more control and spot problems early. And the most common one of all, forgetting to paste the facts. The AI doesn't know your notes, your numbers or your policy. If you don't give them, it will politely make up something that sounds right. That's not a bug in the AI. It's a gap in the brief.
7:09 Recap and try this now
Let's recap. Every good everyday prompt has four ingredients. Who, Aim, Rules, Format. Plus the facts, pasted underneath. When the draft isn't right, fix it with a targeted follow up instead of starting again. Here's your try this now. Pick one task you'll do today, an email, an update, a short guide. Write it with W A R F. Then save the prompt in a note called My prompts, replacing the specific details with brackets, so next time it takes ten seconds. Aim to save five prompts this week. That little library is where the real time savings come from.
Four ingredients
You don't need special tricks to write good prompts. Most everyday tasks go well when you include four ingredients. Remember them as W-A-R-F: Who, Aim, Rules, Format.
- Who: who you are and who the audience is.
- Aim: what you want the AI to produce and why.
- Rules: tone, length, things to include or avoid.
- Format: how the result should look (email, bullets, table, steps).
The template
Who: I'm a [your role]. This is for [audience].
Aim: Please [write / summarize / plan / explain] [what] so that [goal].
Rules: [friendly / formal / simple], [length], include [x], avoid [y].
Format: [email with subject line / 5 bullets / table with columns ...].
My notes: [paste your notes, draft or information]Before and after
Before: "Write a message about the office party."
After:
Who: I'm the team coordinator. This is for our 25-person team.
Aim: Write a short message inviting everyone to our end-of-year dinner so they reply by Thursday.
Rules: warm and inclusive, under 100 words, mention that vegetarian and halal options are available, avoid alcohol references.
Format: a Teams message with a clear call to action.
My notes: Dinner on 18 December, 7pm, at the restaurant near the office. Reply in the poll.The second prompt produces a message you can send with a quick check. The first produces a generic invitation you'd have to rewrite.
Three follow-ups that fix most problems
AI rarely gets everything perfect the first time. Instead of starting over, reply with one of these:
- "Shorter and simpler." Great for emails and announcements.
- "Keep the structure, but change [specific thing]." Keeps what works.
- "Ask me what you need to know to make this better." Lets the AI fill gaps by asking you.
Give examples when style matters
If you want a particular style, paste an example: "Here's a message I wrote last month that my team liked. Match this style." Examples are one of the fastest ways to get output that sounds like you.
Save your best prompts
Create a simple document called "My prompts". Each time a prompt works well, save it with a title like "Team announcement" or "Polite follow-up". Replace the specific details with [brackets] so you can reuse it. Within a month, you'll have a personal toolkit that saves hours.
Worked example
Bilal, a customer service team lead in Karachi, needs to explain a new refund process to his team. His first attempt ("explain refund process") gives a generic explanation of refunds. Using the formula, he adds his role, the audience (new agents), the aim (process refunds correctly in the new system), the rules (simple English, numbered steps, include the approval limit) and pastes the policy text. The result is a clear 8-step guide. He checks it against the policy, fixes one step and shares it.
Sample output check
A good response to the dinner prompt is under 100 words, includes the date, time, place and reply deadline, mentions both dietary options, and ends with a clear action ("Please vote in the poll by Thursday"). If any of these is missing, ask for it specifically rather than regenerating.
Break big tasks into steps
For larger tasks, don't ask for everything at once. Ask for a plan first, then work through it one step at a time: "First, list the sections this guide needs. Wait for my approval before writing." Smaller steps give you more control and better results, and they make it easy to spot where something went wrong.
Hands-on: three before-and-after prompts
Try each "after" prompt in your own assistant and compare it with the "before" version.
1. A status update
Before: Write a project update.
Who: I'm a project coordinator. This is for our regional director, who has two minutes.
Aim: Summarize this week's progress on the Jeddah store fit-out so she knows whether to worry.
Rules: plain English, under 120 words, lead with the overall status (on track / at risk / late), no jargon.
Format: one status line, then three bullets: done, next, risks.
My notes: flooring done; lighting delayed 4 days (supplier); signage approved; opening date still 14 Oct.Typical result: "Status: on track, with one risk. Done: flooring complete, signage approved. Next: lighting installation (now due Thursday). Risk: a further lighting delay would compress the final inspection window; opening still planned for 14 October."
2. A how-to guide for a colleague
Before: Explain how to do expenses.
Who: I'm the finance assistant. This is for new sales staff who have never used our system.
Aim: Explain how to submit an expense claim so it is approved the first time.
Rules: numbered steps, simple English, include the rule that receipts over 50 need a photo, avoid finance jargon.
Format: numbered list of at most 8 steps, then a "common rejections" box.
My notes: [paste the policy text]3. Planning a small event
Before: Plan a team lunch.
Who: I'm a team lead in Manchester. This is for 12 people, including two vegetarians and one person who doesn't drink.
Aim: Give me three lunch options within a budget of 25 pounds per head (illustrative) that I can book this week.
Rules: walking distance from the city center, inclusive options, no chain restaurants.
Format: a table: option, why it fits, what to check when booking.Note the last line of the table request: "what to check when booking". Asking the AI to tell you what to verify is a small trick that pays off in every task.
Let the assistant remember your W and R
If you type the same "Who" and "Rules" every day, store them once: custom instructions in ChatGPT, profile preferences or a Project in Claude, a Gem in Gemini, or saved prompts in Copilot. Then your everyday prompt shrinks to Aim, Format and notes. You'll set this up properly in the "Projects, memory and custom instructions" lesson.
Measure it
A prompt is good when the first answer needs only light edits. Track "rounds to a usable answer" for a week. If most tasks take three or more follow-ups, your Who or Rules are too thin.
Pitfalls
- Leaving out the audience; tone depends on it.
- Asking for "everything" in one go. Break big tasks into steps.
- Forgetting to paste the actual information the AI needs.
Key takeaways
- W-A-R-F: Who (you and the audience), Aim (what and why), Rules (tone, length, include/avoid), Format (shape of the result).
- Always paste the facts the AI needs; it cannot know your notes, numbers or policies otherwise.
- Fix drafts with targeted follow-ups (shorter, keep structure but change X, ask me questions) instead of starting over.
- Save prompts that work, with [brackets] for details, and store your usual Who and Rules in custom instructions or a Project.
Check your understanding
Quick questions to lock in the lesson. They don’t count towards your certificate.
Put it into practice
Rewrite three prompts you use regularly using W-A-R-F and save them in a 'My prompts' document.
Enrol for free to save your progress
Reading is always free. Enrol to keep your place, take the final assessment and earn a verifiable certificate.