Participant Workbook

Raglan AI

Hands-on AI training for people who want to actually use this stuff

Format: Full-day
Date: __________________
Venue: __________________



This workbook is yours to keep. Use it during exercises, scribble in the margins, and take it home as a reference. The workshop platform stays live at tekete.co.nz/ai-workshop — come back anytime.

Raglan AI
Module 01 · What is AI, really?
Module 01

What is AI, really?

Understand what AI can and can't do — in plain language, without the hype.

By the end of this module you'll be able to:
  • Explain how a language model works in one sentence
  • Name three things AI is genuinely useful for — and two where it routinely fails
  • Describe what a "hallucination" is and why it happens
Exercise 1.1
What do you already know?
Before we start — write down one thing you've heard about AI that you're not sure whether to believe. We'll come back to it.
Exercise 1.2
Your first prompt
Open ChatGPT, Claude, or Gemini (whichever you have). Type: "Explain what you are in one sentence, like I'm 12." Write what it says below. Does it sound right to you?
Raglan AI
Module 02 · Prompt Fundamentals
Module 02

Prompt Fundamentals

Learn the four-part formula that turns vague requests into useful results.

By the end you'll be able to:
  • Write a prompt using the RCFF formula (Role · Context · Format · Focus)
  • Explain why specificity changes output quality
  • Improve a weak prompt in under 60 seconds
The RCFF Formula
Role
Act as a...
Context
Here's the situation...
Format
Give me a...
Focus
The key thing is...
Exercise 2.1 — Fix this prompt
Weak → Strong
This prompt is too vague. Rewrite it using RCFF:
Original (weak)"write me an email"
Exercise 2.2 — Write your own
A prompt that matters to your work
Think of one document or message you regularly have to write. Build a prompt using RCFF for it. Test it. What changed?
Raglan AI
Module 03 · Creative AI
Module 03

Creative AI

Use AI to generate images, copy, and creative assets — without needing design skills.

By the end you'll be able to:
  • Write an image prompt that reliably produces what you want
  • Use AI to draft social posts, event copy, and promotional text
  • Understand the copyright questions around AI-generated images
Exercise 3.1 — Image generation
Make a poster concept
Using DALL·E (in ChatGPT Plus) or Adobe Firefly (free), describe an event poster for your organisation. Be specific: colours, mood, subject, style. Write your prompt below and note what the output was like.

What worked / what didn't:
Exercise 3.2 — Copy generation
Write three versions, pick the best
Ask an AI to write a social media caption for your organisation in three different tones (e.g. warm, professional, playful). Which feels right? What would you change?
Raglan AI
Module 04 · Your Terminal Unlocked
Module 04

Your Terminal Unlocked

Open a terminal, navigate your file system, and run tools that only exist on the command line.

By the end of this module you'll be able to:
  • Navigate your file system using cd and ls (or dir on Windows)
  • Install Node.js and run npm to add tools to your computer
  • Explain why developers use the terminal rather than a GUI
Exercise 4.1 — Decode 5 words
Terminal jargon, in your own words
Without looking anything up: write what you think these five words mean. We'll compare with the group. There are no wrong answers — this is a baseline, not a test.
TERM
WHAT I THINK IT MEANS
Terminal
npm
Node.js
API
PATH
Exercise 4.2 — Get one tool running
Document your first install
What tool did you install? What command ran it? What happened? Note any errors you hit and how you fixed them — these notes are gold for next time.
Tool installed:
Command I ran:
What happened / any errors:
Raglan AI
Module 05 · Build Something Real
Module 05

Build Something Real

Use an AI-powered IDE to build a working web page from scratch — something you'd actually use or share.

By the end of this module you'll be able to:
  • Write a clear project spec before you build — what it is, who it's for, what it must do
  • Use Bolt.new, Cursor, or v0 to generate a working first version from a prompt
  • Iterate the build with follow-up prompts to get closer to what you actually want
Exercise 5.1 — Project spec
What are you building?
Before touching the AI: write your spec. A good spec makes the AI's first draft much closer to what you want. Answer these three questions.
What is it? (one sentence)
Who is it for? (be specific)
What must it do? (3–5 bullet points)
Exercise 5.2 — First iteration
What did the AI build, and what's next?
After the first AI build: note what worked and write your next prompt to improve it. The first draft is rarely the final draft — iteration is the skill.
What worked in the first build?
Your next iteration prompt:
Raglan AI
Module 06 · Make It Real
Module 06

Make It Real

Version control your build, connect an AI API, and ship something live with a real URL.

By the end of this module you'll be able to:
  • Create a Git repository, stage files, and write meaningful commit messages
  • Get an API key from Anthropic or Google and integrate it into a build safely
  • Deploy a live URL using GitHub Pages or Cloudflare Pages
Exercise 6.1 — Builder's audit
What needs improving before you'd share this?
Look at your Module 5 build honestly. List three things that still need work. Then pick the one that matters most and fix it before moving on.
1
2
3
The one I'll fix right now:
Exercise 6.2 — Ship it
First commit, and your live URL
Make your first Git commit with an honest message. Then deploy to GitHub Pages or Cloudflare. Write the details below — including the live URL when you have it.
Commit message I used:
Live URL (once deployed):
API I connected (if any) and what it does:
Raglan AI
Module 07 · Advanced Vibe Coding
Module 07

Advanced Vibe Coding

Debug with AI, work across multiple files at once, and push your build further than a single session.

By the end of this module you'll be able to:
  • Run the debugging loop: identify the error → describe it to AI → apply fix → test → repeat
  • Use screenshot-to-code to build UI from visual reference
  • Give AI precise multi-file context so it doesn't break existing work
Exercise 7.1 — The debugging loop
Fix something that wasn't working
Find one thing that's broken or incomplete in your earlier builds. Document the debug loop: what was the problem, what did you tell the AI, what did it do, did it work?
What was broken:
What I told the AI (my exact prompt):
What happened (fixed / partially fixed / still broken):
Exercise 7.2 — Dream build
If you knew you could build anything
No constraints. What would you build if you had unlimited time and no fear of getting stuck? Describe it as a spec — this is the goal you're working towards after today.
Raglan AI
Module 08 · Make Your Own Agent
Module 08

Make Your Own Agent

Understand agent architectures, build a simple research agent, and plan the agent that would genuinely help your work.

By the end of this module you'll be able to:
  • Explain the difference between a "prompt" and an "agent" — what makes something agentic
  • Build a research agent using a system prompt that plans → searches → synthesises
  • Design a real agent specification for your own business or work context
Exercise 8.1 — Research agent output
Run the research agent on a real topic
Use the research agent prompt from Module 8 on a question that actually matters to your work. Document the topic, what it found, and how accurate / useful it was.
Research topic I used:
Key things it found (3 bullets):
Accuracy / usefulness (1–5) and why:
Exercise 8.2 — Your agent roadmap
Plan the agent that would help you most
What would an agent built specifically for your work actually do? Describe the task it handles, what it needs to know, and what output it gives you. This is a real spec you could actually build.
What task does it handle?
What does it need to know? (tools, data, context)
What output does it give you?
Raglan AI
Key Terms & Notes

Key Terms

Large Language Model (LLM)
An AI trained on huge amounts of text to predict the next word. Not a search engine — it generates, not retrieves.
Hallucination
When an AI confidently states something false. Happens because it's predicting, not knowing.
Prompt
The instruction you give to an AI. Quality of output depends heavily on quality of prompt.
Context window
How much text an AI can "remember" in one conversation. If a chat is too long, earlier context may be forgotten.
Temperature
A setting controlling how creative/random the AI's output is. Higher = more creative, lower = more consistent.
Agent
An AI that can take actions — browse the web, run code, send emails — not just respond to text.

Notes

Raglan AI
Tools Reference

Tools to try after today

All of these have free tiers. Start with one or two — don't try to learn everything at once.

ToolBest forAccessCost
ChatGPTGeneral writing, research, summarisingchat.openai.comFree / $20/mo
ClaudeLong documents, careful analysis, codingclaude.aiFree / $20/mo
GeminiGoogle Workspace integration, researchgemini.google.comFree
DALL·E 3Image generation (via ChatGPT)In ChatGPT PlusNeeds Plus
Adobe FireflyCommercial-safe image generationfirefly.adobe.comFree credits
Canva AIDesign + AI image combocanva.comFree tier
Otter.aiMeeting transcription & summariesotter.aiFree tier
Notion AIDocs, notes, databases + AI assistnotion.soAdd-on
PerplexityAI search with citationsperplexity.aiFree
Google AI StudioAPI access, experiment with Geminiaistudio.google.comFree

Prompt starters to keep on your desk

Summarise a document
"Summarise the following in 5 bullet points for a non-expert audience: [paste text]"
Write an email
"As a [role], write a [tone] email to [recipient] about [topic]. Keep it under 150 words."
Fix my writing
"Improve the clarity and tone of this text without changing its meaning: [paste text]"
Generate ideas
"Give me 10 ideas for [goal]. I'm working with [constraints]. Be specific."
Create an image
"[Subject], [setting], [lighting], [style/mood], [colour palette], photorealistic / illustration"
Raglan AI
What Next

What to do this week

The biggest risk after a workshop is going home, meaning to try things, and never quite getting started. Here's a concrete 5-day plan.

Day 1 — Today
Do one real thing
Use AI to draft something you'd normally spend 20+ minutes on — an email, a social post, a report intro. It doesn't have to be perfect. Just do it once.
Day 2
Share it with someone
Show a colleague what you made. Explain the prompt you used. Teaching it is the best way to cement it.
Day 3–4
Explore one more module online
Head to tekete.co.nz/ai-workshop and try a module you didn't get to today. The branch modules are shorter and focused.
Day 5
Identify your "AI routine"
Write down the one thing in your work week that AI could genuinely save you 30+ minutes on. Make that your practice ground for the next month.
Stay in touch

The workshop platform is always live at tekete.co.nz/ai-workshop

Questions? tobias@tekete.co.nz — I'm happy to help.