Module 02

AI for Your Work,
Right Now

No terminal. No code. Move from treating AI like a search engine to treating it like a brilliant, literal-minded intern.

⏱ ~40 min self-directed 🌐 Browser only ✍️ Live prompt exercises
Your progress0%

Read the content Β· complete the tests = 100%

✍️ The anatomy of a prompt

The biggest gap between bad AI results and great AI results isn't the model you use β€” it's how you ask. Most people write one vague sentence. To an AI, vague means "guess what I want based on statistical internet averages." You don't want statistical average.

❌ The Typist (Weak)
Write an Instagram caption for my cafΓ© selling winter soup. Make it sound good.
βœ… The Manager (Strong)
You are an expert social media manager for a small beachside cafΓ©. Our vibe is unpretentious, local, and warm. Write 3 options for an Instagram caption announcing our new winter seafood chowder. Do not use emojis in the middle of sentences. Keep each under 3 sentences. No hashtags.
Role Context Task Constraints
Exercise 2.1 β€” Live Test
The Prompt Strength Tester

Pick a real task from your job (an email, report, or social post). Write the prompt below. Try to hit all four markers to unlock the 'Prompt Engineer' achievement.

Strength Score Analyzing...
Detail (>20 words)
Role assigned ("act as", "you are")
Format specified ("list", "email", "table")
Constraints used ("don't", "keep it under", "limit")

🧠 The "Messy Brain Dump"

A huge psychological block for beginners is assuming they have to write polite, perfectly formatted instructions for the AI. You don't.

AI is brilliant at synthesis. You can paste a chaotic, stream-of-consciousness wall of text, broken meeting notes, and half-formed thoughts, and simply say: "Synthesize this mess into a structured 3-point email."

Show, Don't Tell (Few-Shot Prompting): Instead of giving the AI 10 paragraphs of rules about how you write, just paste one of your previous emails and say, "Mirror this tone exactly." It's infinitely faster and produces far better results.

πŸ₯ Humanising & Localising

If you don't constrain the AI, it defaults to what we call "American Corporate Robot." It uses words like "Delve", "Leverage", and "Transformative." It spells 'colour' wrong. It sounds intensely un-Kiwi.

The NZ Localization Block
Tone instructions: Use British/New Zealand English spelling (e.g. organise, colour). Avoid overly enthusiastic corporate jargon, acronyms, and words like "delve" or "leverage". Keep the tone grounded, informal, and kiwi. Incorporate extremely light, natural te reo Māori greetings where appropriate (e.g., Kia ora, whānau) but do not overdo it.

Keep that snippet on your desktop. Paste it at the bottom of any prompt when you need something to sound like it was actually written in Aotearoa.

πŸ”‘ Secret Codewords & Tags

The smartest models (like Claude 3.7 or DeepSeek) deeply respect "XML tagging" to break down complex instructions, and explicit commands to slow them down.

Exercise 2.2 β€” The Power User Toolkit
Codewords & Meta-Prompting

The ultimate power move is realising you aren't talking to a search engine. You can command it to think, critique, and interrogate you.

Open your AI of choice right now. Copy one of these three structural prompts and test it:

1. The Socratic Interrogator
I need to write a proposal for a new community garden project, but I don't know where to start. You are an expert strategist. Before you write anything, ask me 3 highly specific, challenging questions that will pull the best information out of me. Ask them one at a time.
2. The Adversarial Critic
I am planning to launch a weekend market stall selling handmade ceramics. Act as a ruthless business critic. Tear this idea apart. Give me the 4 biggest reasons this will fail, and then provide exactly 1 structural solution for each. Don't be polite.
3. The Structured Thinker
Write a polite but firm email to a supplier who is 3 weeks late on delivering our store's packaging.

<thinking>
First, outline your goal and the tone you want to strike.
</thinking>

Now, write the final email below.

Notice the shift: You are no longer commanding a typist. You have hired a consultant.

Module 2 done πŸŽ‰

You now know how to prompt like an engineer. In Module 3, we explore what AI can do for creative work β€” images, music, writing, video β€” before we touch the terminal in Module 4.

Want to go further on prompting? This branch goes deep: