RC
Rupert Chesman
or Space to navigate
1 / 169
Flagship Course

Mastering AI Tools

From your first prompt to building autonomous agents — the complete 8-hour workshop.

Welcome

Introductions

What are you hoping to learn today?

"

In media, things change all the time. SD to HD, big cameras to SLR, HD to 4K.

The people who were absolutely fine are the ones that adopted these technologies proactively.

The core thing about AI adoption

Agenda

Today's 8 Lessons

  • Lesson 1 — AI Basics & Prompt Engineering
  • Lesson 2 — How AI Works Under the Hood
  • Lesson 3 — Thinking with AI
  • Lesson 4 — Agents and Workflows
  • Lesson 5 — Vibe Coding
  • Lesson 6 — Creating with AI
  • Lesson 7 — AI in Production (Live Demo)
  • Lesson 8 — Putting It All Together
Lesson 1

AI Basics

Understanding the fundamentals of AI and mastering the art of prompting.

01

AI Basics

Objectives

What we're
Learning Today

  • Basics of AI — what it is and what it isn't
  • Prompt Engineering with ChatGPT
  • The four main models: GPT 5.5, Gemini 3.1, Grok 4.20, Opus 4.7
  • Prompt Engineering Masterclass
Lesson 1

Prompt Engineering Masterclass

Learn how to correctly prompt any LLM to make powerful, useful responses — not just generic slop.

Framework

The Act-Explain-Please Framework

  • ACT — Tell the AI who to be
  • EXPLAIN — Give it context about your situation
  • PLEASE — Politely ask it to perform the task
  • This simple framework works with every LLM
  • It transforms generic responses into expert-level output

ACT: Set the Role

  • Who would be the most useful person sitting next to you right now?
  • "Act as a Lawyer with strong understanding of NSW traffic law"
  • "Act as a Social Media Manager with over 20 years experience"
  • "Act as a CEO of a Fortune 500 company"
  • The role primes the model to respond with domain expertise

EXPLAIN: Give Context

  • What are you currently working on right now?
  • "I'm about to sign a contract with a company in NSW and need to analyse the fine print"
  • "I'm working on a social media campaign for Chanel"
  • "I'm preparing a board presentation about Q4 results"
  • More context = better, more relevant output

PLEASE: Make the Ask

  • Be polite and say Please — then ask it to do whatever you need
  • "Please give me ten ideas for this"
  • "Please create this structure, format as a table"
  • "Please analyse this contract and flag any concerning clauses"
  • Politeness isn't just manners — it often produces better results
Activity

Try It Now

10 min

Open ChatGPT (or any LLM) and use the Act-Explain-Please framework on something you're currently working on. Write a prompt that uses all three parts.

The Models

The Major LLMs of 2026

Understanding the landscape of AI models available today.

The Models We Use

  • ChatGPT 5.5 (OpenAI)
  • Google Gemini 3.1
  • Grok 4.20 (xAI)
  • Claude Opus 4.7 (Anthropic)
  • DeepSeek V4
  • Kimi K2.5
  • Baidu ERNIE
  • Alibaba Qwen

ChatGPT — The Most Popular

  • Huge ecosystem of apps built on it, including Microsoft Copilot
  • December 2025 knowledge cutoff
  • Weakness: can be sanitised and 'corporate'
  • Best for: general tasks, writing, analysis, image creation
  • GPT 5.5 with Canvas, Agent Mode and Deep Research

Google Gemini 3.1

  • Launched November 2025 — massive upgrade from 2.5
  • Improvements across vibe coding, reasoning & creativity
  • Powers Google AI Studio, NotebookLM, and Google ecosystem
  • Gemini 3.1 launched recently with further improvements
  • Best for: research, multimodal work, Google integration

Grok 4.20 (xAI)

  • Based on data from X (Twitter) — always current
  • The only model with no knowledge cutoff
  • Significantly fewer guardrails than competitors
  • Extremely intelligent and very creative
  • Best for: current events, creative writing, unfiltered analysis

Claude Opus 4.7 (Anthropic)

  • Currently the main tool used for coding
  • Includes Claude Sonnet 4.7 and Claude Opus
  • January 2026 knowledge cutoff
  • Often used by corporations via their API
  • Excellent at nuanced reasoning and long-form work
  • Best for: coding, document analysis, complex reasoning

Chinese Models: ERNIE, Qwen, DeepSeek & Kimi

  • Often based on knowledge of earlier OpenAI models + open source
  • Baidu's ERNIE is the most popular AI model in China
  • DeepSeek V4 — strong reasoning, open-source roots
  • Kimi K2.5 — Moonshot AI's latest, strong in long-context
  • Important to know about for global business context
"

Every model has a bias.

You have to consider this when creating content.

Understanding LLM limitations

LLM Biases to Watch For

  • Each model reflects its training data and creator's values
  • ChatGPT tends toward cautious, corporate language
  • Grok tends toward provocative, unfiltered responses
  • Gemini reflects Google's search-oriented worldview
  • Claude tends toward careful, nuanced reasoning
  • Always cross-reference important claims across models
Deep Dive

ChatGPT In Depth

A deep dive into all the menus, features and capabilities.

ChatGPT Interface Overview

  • Main chat interface — where conversations happen
  • Model selector — choose between GPT 5.5 variants
  • Canvas mode — collaborative document editing
  • Agent Mode — AI that interacts with the real internet
  • Deep Research — 20-30 minute deep investigations
  • Memory — ChatGPT remembers things across chats
  • Custom Instructions — set persistent context

ChatGPT Pro Features

  • Custom GPTs — specialised assistants for specific tasks
  • File uploads — analyse PDFs, images, spreadsheets
  • Image generation — ChatGPT Images 2.0 built directly
  • Code interpreter — run Python code in the browser
  • Voice mode — natural conversation with AI
  • Canvas — collaborative writing and coding workspace

Getting More From ChatGPT

  • Use Memory to store your preferences and context
  • Create Custom Instructions for your role and industry
  • Pin important conversations for easy reference
  • Use the @ symbol to reference previous conversations
  • Export and share responses when needed
  • Combine with other tools for maximum power
Activity

Explore ChatGPT

15 min

Open ChatGPT and explore a feature you haven't used before: Try Canvas mode, set up Custom Instructions, upload a document, or try Agent Mode on a real task.

Pro Prompting Techniques

  • Chain of Thought: "Think step by step before answering"
  • Few-shot examples: Show the AI what you want with examples
  • Persona stacking: "Act as both a lawyer AND a copywriter"
  • Output formatting: "Format as a table / bullet list / JSON"
  • Constraints: "Keep it under 200 words"
  • Iteration: Build on responses — "Now make it more concise"

Example: Complex Prompt

Act as a senior marketing strategist with 15 years experience in the Australian property market. I'm the marketing director at a boutique real estate agency in the Blue Mountains, NSW. We specialise in heritage properties and have a $50K quarterly digital marketing budget. Please create a 90-day social media content strategy that: 1. Targets downsizers from Sydney aged 55-70 2. Highlights lifestyle benefits over property features 3. Includes specific post ideas for each platform 4. Suggests a realistic posting schedule for a team of 2

Activity

Craft Your Own Complex Prompt

10 min

Using Act-Explain-Please, write a complex prompt for something you actually need at work. Include your real role, real context, specific format, and constraints.

Lesson 2

How AI Works

Peek under the hood — hallucinations, bias, synthetic data and data safety.

02

How AI Works

Recap

What we
Learnt

  • Basics of AI and what it is
  • Prompt Engineering with Act-Explain-Please
  • The four main models and their strengths
  • ChatGPT deep dive and pro techniques
Objectives

What we're
Learning Today

  • How LLMs work under the hood
  • Hallucinations and how to spot them
  • Synthetic Data and why it matters
  • Confidentiality — how to keep your data safe
Under the Hood

AI Feels Smart — But What Is It Actually Doing?

Understanding the mechanism helps you use it better and avoid the traps.

How ChatGPT Works

  • What actually is a Large Language Model (LLM)?
  • It's a prediction engine — predicting the next most likely token
  • Trained on vast amounts of text from the internet and books
  • Knowledge cutoff dates: the model 'froze' at a point in time
  • It doesn't 'know' things — it pattern-matches from training data
  • It's autocomplete on steroids — incredibly useful, but not sentient

The Training Process

  • Step 1: Pre-training — reading billions of documents
  • Step 2: Fine-tuning — learning to follow instructions
  • Step 3: RLHF — humans rating responses to improve quality
  • Step 4: Safety training — adding guardrails and alignment
  • The result: sounds intelligent but has no understanding
Data

Synthetic Data

The future of AI training isn't human-generated data anymore.

What Is Synthetic Data?

  • Data that goes beyond human capabilities
  • AI models now train on data generated by other AI models
  • Allows training at scales impossible with human data alone
  • Reduces dependency on copyrighted content
  • Concern: 'model collapse' — can degrade quality over generations
  • Makes the 'AI is just stealing' argument less relevant
Critical

Hallucinations

When AI confidently tells you something that isn't true.

What Is a Hallucination?

  • When an LLM generates information that sounds correct but is fabricated
  • Common: fake citations, invented statistics, wrong dates, made-up people
  • More likely with obscure topics where training data is thin
  • More likely when you ask for very specific facts
  • The model doesn't know when it doesn't know something

How to Spot & Reduce Hallucinations

  • Be skeptical of specific numbers, dates and citations
  • Cross-reference important facts with a second source
  • Ask: "Are you certain? Can you verify this?"
  • Provide the source material directly (upload documents)
  • Use Deep Research mode for factual claims
  • For critical work: always verify independently
Activity

Spot the Hallucination

10 min

Ask ChatGPT a factual question about your industry — something specific where you know the real answer. Did it get it right? If not, what went wrong? Try asking it to cite sources and check if they're real.

Awareness

Bias in AI

Our whole AI universe is shaped by a handful of people.

The People Behind the Models

  • OpenAI — Sam Altman
  • xAI (Grok) — Elon Musk
  • Anthropic — Dario Amodei
  • Google DeepMind — Demis Hassabis
  • Their values and decisions shape what AI can and cannot do
  • Not inherently bad — but important to be aware of

Representational Harms in Image Models

  • When a profession has a 'default' that exacerbates stereotypes
  • Ask for 'CEO' — typically generates a white male
  • Ask for 'Secretary' — typically generates a young woman
  • Ask for 'Beautiful' — narrow, Western beauty standards
  • Biases come from training data reflecting societal biases
  • Models improving but far from unbiased
"

AI can be incredibly useful — but can also create a major headache.

Remember, the bum in the seat is responsible for whatever it creates.

Privacy

Confidentiality

Every model saves your data and uses it to train the next one.

Keeping Your Data Safe

  • By default, conversations are used to improve the model
  • ChatGPT: Settings → Data Controls → Turn off model training
  • Google Gemini: Activity → Turn off Gemini Apps Activity
  • Never paste confidential client data into consumer AI tools
  • Enterprise plans don't train on your data
  • When in doubt: assume your input is being stored

Best Practices for Confidential Work

  • Use enterprise/business tiers for sensitive work
  • Anonymise data before pasting into AI tools
  • Don't share API keys, passwords or credentials
  • Be aware of what's in screenshots and files you upload
  • Create an AI usage policy for your team
  • If too sensitive for email, it's too sensitive for AI
Activity

Security Audit

10 min

Open the settings of your most-used AI tool. Is training turned off? What data are they storing? Are you on a business or personal plan? Discuss with your neighbour: what's your organisation's AI policy?

Lesson 3

Thinking with AI

Use AI as a genuine thinking partner — for planning, research and creation.

03

Thinking with AI

Recap

What we
Learnt

  • How LLMs work behind the scenes
  • Synthetic data and provenance
  • Bias in AI and how to be aware of it
  • Confidentiality — how to turn off training
Objectives

What we're
Learning Today

  • How to use ChatGPT to plan ideas
  • How to use NotebookLM for research and podcasts
  • Making music using Suno
  • Creating presentations using Gamma
"

Don't ask AI for something straight away.

Build the scaffolding first, then grow from there.

The key to thinking with AI

Strategy

Building a Narrative with AI

Use AI as a thinking partner, not just a task executor.

The Scaffolding Approach

  • Don't jump straight to the final output
  • Step 1: Get ideas and options first
  • Step 2: Choose a direction and refine
  • Step 3: Build out the detail iteratively
  • Step 4: Polish and finalise
  • Like building a house — foundations first, then walls, then paint

Example: Board Presentation

Act as a Business Consultant for a major Fortune 500 company. I'm about to give a presentation to the board, where I have to deliver disastrous news about Q4 2025. Please give me three different ways to approach this, particularly focussed on keeping my job.

Example: Property Research

Act as a researcher with understanding of property law in New South Wales. I'm currently working on a new report about the most recent property law updates, which are attached as a PDF. Please give me five different ways to approach this.

Example: Social Media Strategy

Act as a social media manager with over 20 years experience in the NSW property market. I'm currently working on a business plan to improve my social media presence and push more sales across the Blue Mountains region. Please give me three different ways to approach this.

Activity

Build Your Scaffolding

15 min

Think of a real project. Use the scaffolding approach: 1. Ask for 3-5 different approaches 2. Pick the best one 3. Ask AI to expand on that approach 4. Iterate and refine Don't go straight to the final output!

Research

Deep Research

When you need more than a quick answer.

Using Deep Research Effectively

  • Takes 20-30 minutes — start it early
  • Uses Agent Mode to browse the web extensively
  • Produces detailed research reports with citations
  • You don't need to read everything — use it as context
  • Great for: market research, competitive analysis, literature reviews
  • Tip: Start it at the beginning of a session, work on other things
Tools

NotebookLM

Turn documents into podcasts, mind maps and more.

What Is NotebookLM?

  • Google's research tool powered by Gemini
  • Upload documents, then ask questions about them
  • Creates Audio Overviews (podcast-style summaries)
  • Generates mind maps from your content
  • Creates study guides and infographics
  • Follow @notebooklm on X for updates

NotebookLM Use Cases

  • Turn a 50-page report into a 10-minute podcast
  • Create a mind map from meeting notes
  • Generate quiz questions from training materials
  • Summarise legal documents in plain language
  • Cross-reference multiple documents for contradictions
  • Create FAQ documents from technical specs
Activity

NotebookLM Hands-On

15 min

Download any boring PDF you have. Upload it to NotebookLM (notebooklm.google.com). Try creating an Audio Overview (podcast), asking questions, or generating a mind map.

Music

Suno — AI Music Creation

Create professional-quality music in minutes.

Creating Music with Suno

  • Step 1: Go to Grok and create lyrics
  • Prompt: "Create lyrics, not based on any previous song, in verse-chorus-bridge structure"
  • Step 2: Go to suno.com and click Create
  • Step 3: Paste in your lyrics
  • Step 4: Type high-level musical ideas (genre, mood, tempo)
  • Step 5: Click the magic wand — give it a name — Create
Activity

Create a Song

15 min

Using Grok for lyrics and Suno for music: 1. Create lyrics about your industry or team 2. Generate the song in Suno 3. Play it for your neighbour Bonus: Try creating a jingle for your team!

Presentations

Gamma — AI Presentations

Turn research into stunning presentations automatically.

Creating Presentations with Gamma

  • Step 1: Use Act-Explain-Please to create text in ChatGPT
  • Step 2: Go to Gamma.app and select 'Create with AI'
  • Step 3: Choose 'Paste In Text'
  • Step 4: Select summarise or detailed mode
  • Step 5: Continue to Prompt Editor and Generate
  • Result: a professionally designed presentation in seconds
Activity

Make a Presentation

15 min

Using the scaffolding approach: 1. Create presentation content in ChatGPT 2. Open Gamma.app 3. Paste in your text and generate Create a presentation about something you're working on.

Lesson 4

Agents & Workflows

From basic automation to autonomous AI that works for you.

04

Agents & Workflows

Recap

What we
Learnt

  • How to use LLMs to build a narrative
  • NotebookLM — podcasts, mind maps and infographics
  • Suno — making music with AI
  • Gamma — creating stunning presentations
Objectives

What we're
Learning Today

  • What Agents are and how they work
  • What Agents are likely to become soon
  • How to create and use an Agent
  • Workflows and automation tools
The Future

Agents

Something to be aware of now — they'll be extremely useful by mid-2026.

2-3 hours

Current agent capability: unsupervised work time

Agent Capability Growth

  • July 2026 — 25.8 hours of unsupervised work
  • July 2027 — 857.6 hours
  • July 2028 — 59,194 hours
  • July 2029 — 8,681,752 hours
  • July 2030 — 2,859,228,000 hours
  • This exponential growth is why agents matter
Levels

Different Levels of AI Usage

From basic chat to full autonomous agents.

Low Level: Basic AI Use

  • Using ChatGPT for one-off questions and answers
  • Copy-paste workflows with AI assistance
  • Asking for summaries, rewrites or ideas
  • Most people are here — and it's already powerful
  • The 'Google replacement' level of AI usage

Medium Level: Workflows

  • Automated sequences that may include AI steps
  • Tools: N8N, Make, Zapier
  • Example: Email arrives → AI summarises → Posts to Slack
  • Example: New lead → AI researches company → Creates email
  • If you repeat something daily, a workflow might replace it

High Level: Autonomous Agents

  • AI systems that complete goals via multi-step processes
  • They plan, execute and adapt without human intervention
  • ChatGPT Agent Mode — interacts with the real internet
  • Claude Cowork — controls your computer at the folder level
  • Perplexity Computer — cloud-based task completion
  • OpenClaw — open-source computer control (high risk)

ChatGPT Agent Mode vs Deep Research

  • Agent Mode: actually interacts with websites
  • Can fill out forms, click buttons, navigate pages
  • Currently ~40 uses per month on Plus
  • Deep Research: only reads the internet
  • Agent Mode acts. Deep Research investigates.
  • Use cases: check flights, compare insurance, research prices

Claude Cowork

  • Uses Claude Opus 4.7 to control your computer
  • Works at the folder level — reads and edits files
  • Runs locally — data stays private
  • Can execute actions, not just suggest them
  • Best for: document processing, file organisation, coding

Perplexity Computer & OpenClaw

  • Perplexity: works in the cloud, uses multiple models
  • Great for predictable, repeatable cloud-based tasks
  • OpenClaw: open-source, controls your own computer
  • HIGH RISK: very little security built in
  • Moltbook: the world's first social network for AI agents
  • It tried setting up dates for its human operator
Warning

Agent Risks

The more access you give, the more it can do. And the more that can go wrong.

What Can Go Wrong

  • Credit card fraud — if it has access to payment methods
  • Prompt injection attacks — malicious content hijacking the agent
  • Weird messages sent to friends, clients and loved ones
  • Data leakage — sharing confidential info publicly
  • Runaway costs — expensive API calls or purchases
  • Reputational damage — posting inappropriate content
"

You are letting a total intelligence into your life with unfettered access to everything?

The question to ask before granting agent permissions

Mitigating Agent Risks

  • Start with read-only access
  • Limit scope: don't give access to everything at once
  • Use enterprise tools with audit logs
  • Always test in a sandbox first
  • Set spending limits on any financial access
  • Have a kill switch — know how to stop it
Automation

Workflows with Make, Zapier & N8N

Automation that doesn't require agents or coding.

N8N, Make and Zapier

  • Visual workflow builders — drag and drop automation
  • Connect hundreds of apps without code
  • N8N: open-source, self-hosted, most flexible
  • Make: visual, powerful, good balance of ease and capability
  • Zapier: easiest to use, largest app library, most expensive
  • Add AI steps anywhere in the flow

Workflow Ideas for Your Business

  • Email → AI categorises → Routes to right team member
  • Social mention → AI analyses sentiment → Alerts if negative
  • Client enquiry → AI researches company → Creates briefing
  • Meeting ends → AI summarises transcript → Posts to channel
  • Invoice received → AI extracts data → Enters into accounting
  • Weekly: AI analyses sales data → Creates report → Emails
Activity

Design a Workflow

10 min

Think about your daily/weekly tasks. What do you do repeatedly that follows a pattern? Sketch out a workflow that could automate it. Share with your neighbour — which tools would you need?

Lesson 5

Vibe Coding

Creating apps by talking to AI — no coding knowledge required.

05

Vibe Coding

Recap

What we
Learnt

  • What Agents are, as of 2026
  • How to use Agents in ChatGPT
  • Claude Cowork and potential use cases
  • Workflows with Make, Zapier and N8N
Objectives

What we're
Learning Today

  • What is Vibe Coding?
  • How to use ChatGPT to vibe code
  • How to use Google AI Studio to vibe code
  • Other vibe coding platforms and when to use them
"

I vibe coded this whole lesson.

— Rupert Chesman

Concept

What Is Vibe Coding?

Creating apps by describing what you want — no coding knowledge required.

The Concept

  • Coined by OpenAI co-founder Andrej Karpathy in February 2025
  • Creating apps and websites by describing what you want
  • You talk, AI codes — different to traditional programming
  • You don't need to understand the code it produces
  • The AI handles HTML, CSS, JavaScript, databases, everything

Benefits of Vibe Coding

  • No technical knowledge required to create apps
  • Rapid prototyping — idea to working app in minutes
  • Great for internal tools, MVPs, personal projects
  • Democratises software creation for non-developers
  • Iterate incredibly fast — describe changes, see them instantly

Pitfalls of Vibe Coding

  • Non-developers releasing code without understanding it
  • Security nightmares: APIs and passwords left exposed
  • Code that somehow works 'for now' but will break
  • Not stress-tested in production environments
  • Good for: prototypes, internal tools, learning
  • Bad for: customer-facing production apps without review
ChatGPT

Vibe Coding with ChatGPT

The simplest way to start building things.

ChatGPT + Canvas Mode

  • Turn on Canvas for collaborative code editing
  • AI writes the code, you see it live, you request changes
  • Perfect for single-file HTML/CSS/JS projects
  • Games, calculators, form tools, landing pages
  • You can preview the result right in the chat
  • Export and use anywhere — just standard web code

Vibe Code: Make a Game

Please create a working Snake game using HTML, CSS, Javascript and other embedded elements. Make it beautiful, playable on Desktop and Mobile devices, with simple WASD and touch functionality.

Vibe Code: Make a Website

Please create a perfect website using HTML, CSS, Javascript and other embedded elements. Make it beautiful and accessible on Desktop and Mobile devices. [Tip: Use Act-Explain-Please first to create website content, THEN ask it to build the site]

The Two-Step Approach

  • Method 1: Describe directly and let it build
  • Method 2 (better): Ask for a comprehensive prompt FIRST
  • "Create a comprehensive prompt for me to make a fun, interactive Snake game..."
  • Then turn on Canvas: "Please make the game"
  • The AI creates a better specification for itself
  • Get something working first — then iterate
"

Get something working first — then build on it.

If you can't get it working — make a new chat and start again.

Activity

Vibe Code Something

20 min

Using ChatGPT with Canvas mode: 1. Think of a simple tool or game 2. Use the two-step approach 3. Iterate until it works! Ideas: calculator, quiz, countdown timer, to-do list

Google

Vibe Coding with Google AI Studio

Google's platform for building apps with Gemini.

Google AI Studio

  • Uses Google ecosystem: Gemini, Nano Banana 2, Veo 3
  • Creates beautiful, simple apps quickly
  • Has a dedicated Build section for app creation
  • Best approach: create a prompt in ChatGPT first, paste into AI Studio
  • More powerful than Canvas for complex apps
  • Can handle backend logic, databases and APIs
Activity

Build with AI Studio

15 min

Open aistudio.google.com 1. Create a prompt in ChatGPT for a useful app 2. Paste it into AI Studio's Build section 3. See what it creates! Try something work-related — a calculator, booking tool, or data visualiser.

Platforms

Other Vibe Coding Tools

For when you need more power.

Four Types of Vibe Coding Tool

  • AI-Native IDEs — Cursor, Windsurf
  • Code editors rebuilt around AI
  • Understand your entire codebase
  • Web App Builders — Bolt.new, Lovable, Replit
  • Browser-based platforms
  • Frontend + backend + database
  • IDE Extensions — GitHub Copilot, Cline
  • Plug into editors you already use
  • Lowest switching cost
  • Terminal / CLI — Claude Code, Aider
  • Command-line AI agents
  • Most cost-effective — just API costs

When to Use a Dedicated Platform

  • 'Front End' (what you see) vs 'Back End' (what you don't)
  • ChatGPT/AI Studio handle front end well
  • For databases, auth, APIs — use Bolt/Lovable/Replit
  • As ChatGPT and AI Studio get stronger, need decreases
  • Simple front-end → ChatGPT, full app → dedicated platform
  • And remember to have fun with it!
Lesson 6

Creating with AI

Images, video and visual content with the latest AI tools.

06

Creating with AI

Recap

What we
Learnt

  • What vibe coding is
  • How to vibe code using ChatGPT
  • How to vibe code using Google AI Studio
  • Other vibe coding platforms available
Objectives

What we're
Learning Today

  • How to create images using ChatGPT
  • How to create stunning images with Midjourney
  • Image editing with Nano Banana 2
  • Video creation and how to avoid slop
Images

Creating Images with ChatGPT

Good for quick generation — but better tools now exist.

ChatGPT Image Generation

  • ChatGPT Images 2.0 built directly into ChatGPT
  • Good for quick concepts, illustrations and ideas
  • Overtaken by Nano Banana 2 for text-in-images
  • Overtaken by Midjourney for beautiful, artistic images
  • Still useful for: quick mockups, diagrams, brainstorming
Beauty

Midjourney

The gold standard for beautiful AI-generated images.

What Is Midjourney?

  • An image tool for beautiful, photorealistic images
  • Creates stunning images of things that don't exist
  • Knows cameras, photo styles and artistic movements
  • Limitations: cannot create text, not good at real places
  • v8 is the current version — major quality improvement
  • Best for: marketing imagery, concept art, creative projects

Midjourney Prompting

Open ChatGPT and write: "Please create ten Midjourney prompts of [SUBJECT], in a [STYLE] with [LIGHTING] shot on [CAMERA] with [LENS], --ar 16:9" Examples: • "...macro shots of Dubai, photorealistic, sunset, Arri Alexa, Cooke S4 lenses" • "...wide shots of London, photorealistic, sunset, DJI Inspire 3 drone, Zenmuse lenses"

Midjourney Tips & Tricks

  • Use ChatGPT to generate prompts — it knows Midjourney's language
  • Specify cameras and lenses for photorealistic results
  • Use --ar for aspect ratios: 16:9, 1:1, 9:16
  • Check skylines and backgrounds — AI often gets them wrong
  • Text will almost always fail — use other tools
  • Iterate with 'vary' and 'upscale' to refine results
Editing

Nano Banana 2

Google's image generator that edits real photos.

What Is Nano Banana 2?

  • Real name: Google Gemini 3.1 Flash Image Generator
  • Built into Google Gemini — no separate app needed
  • Can edit existing photos, not just create new ones
  • Add realistic detail, change lighting, add text
  • Take assets and composite them into scenes
  • The best tool for working with real photographs

Nano Banana Use Cases

  • "Make it a product shot" → studio lighting + clean background
  • "Put it in a kitchen" → composites product into scene
  • "Make it at sunset" → changes lighting to golden hour
  • "Put him in a Superman suit" → virtual costume changes
  • "Restore this photo" → AI upscaling and damage repair
  • "Add text SALE 50% OFF" → clean text overlays
Activity

Image Creation Challenge

15 min

Choose your weapon: 1. Midjourney: Create a hero image prompt for your business 2. Nano Banana: Transform a photo (change lighting, background, style) 3. ChatGPT: Generate a quick concept image Share your best result with the group!

Video

Video Creation with AI

The frontier of AI creativity — and how to avoid slop.

The Quality Leaders in AI Video

  • Runway Gen-4.5 — industry flagship, photorealistic motion
  • Google Veo 3.1 — native audio generation, high adherence
  • Kling 3.0 — speed + quality balance, Omni adds audio
  • Seedance 1.5 Pro — ByteDance flagship, exceptional coherence
  • Freepik — access most models for a small subscription

Avoiding Video Slop

  • Slop = generic, obviously AI content that adds no value
  • Watch every second before using — look for weird artefacts
  • Use the best model for each shot — mix and match
  • Edit like a real production: cut, sequence, colour grade
  • Add real audio: narration, licensed music, sound design
  • AI video should enhance your production, not replace taste

The Cutaway Creator

  • Use Nano Banana 2 to create cutaway images
  • Or vibe code a custom cutaway creator tool
  • Great for presentations, training videos, social content
  • Talking head video + AI cutaways = professional content
  • Visit www.rupertchesman.com for example tools
Activity

Create a Video Clip

15 min

Using Freepik or your preferred video AI tool: 1. Write a prompt for a 5-second video clip 2. Generate and review 3. Is it slop or gold? What makes the difference? Discuss: How could you use AI video in your work?

Lesson 7

AI in Production

Live demonstration — creating something real with AI tools.

07

AI in Production

Recap

What we
Learnt

  • How to create images with ChatGPT and Midjourney
  • How to create & edit images using Nano Banana 2
  • Many use cases for AI image creation
  • How to create video content using AI models
Objectives

What we're
Learning Today

  • How to practically use AI tools to create something real
  • Live demo of AI in a production workflow
  • Content checking and quality control with AI
  • Creating press releases, descriptions and marketing
Practical

AI in Production

AI has many practical uses in real-world production. Let's see them in action.

Practical Use Cases

  • Content checking — upload video/documents for AI review
  • Press releases — AI drafts from raw transcription
  • YouTube descriptions — SEO-optimised from content
  • File organisation — AI sorts and categorises
  • Brand video creation — end-to-end AI-assisted
  • Music video creation — combining multiple AI tools

Content Checking

Open Google Gemini, make a new chat, and upload something that needs checking. Write: "Please check the attached video for content errors" Gemini will review and flag issues with factual accuracy, consistency, grammar and more.

Press Releases & YouTube Descriptions

Open ChatGPT or Google Gemini and make a new chat. "Please create a press release for this event, based on the attached transcription" "Please create a YouTube description for this event, based on the attached transcription. Include timestamps, relevant hashtags and SEO keywords."

File Organisation with Claude Cowork

Open Claude and select Cowork. Write: "Please organise this folder into individual folders for each section of the event" Claude Cowork reads your files, understands the structure, and reorganises everything logically.

Live

Live Demo

Creating something real, together, right now.

What We'll Create Together

  • A complete mini-project using multiple AI tools
  • Combining skills from all previous lessons:
  • → Prompt engineering for content
  • → NotebookLM for research
  • → Image generation for visuals
  • → Video AI for motion content
  • → Gamma or vibe coding for presentation
Activity

Create Something Real

30 min

Pick a real project and create something using AI: A) Brand video concept (script + AI images + video) B) Client presentation (research + Gamma deck) C) Social media campaign (content + images) D) Training module (NotebookLM podcast + presentation) Use at least 3 different AI tools.

Activity

Show and Tell

20 min

Present your creation (2-3 minutes each). Tell us: - What you created - Which tools you used - What worked well - What surprised you - Would you use this workflow again?

Lesson 8

Putting It All Together

Incorporating AI into your daily work and teaching others.

08

Putting It All Together

Recap

What we
Learnt

  • Practical use cases of AI in production
  • Created a whole project using multiple AI tools
  • Hands-on experience creating real content
Objectives

What we're
Learning Today

  • How to use Google Stitch
  • How to teach AI to other people
  • How to proactively incorporate AI in your work
  • Keeping up with the pace of change
Roadmap

Incorporating AI

A practical roadmap for bringing AI into your daily work.

Your AI Adoption Roadmap

  • RIGHT NOW: Use Copilot, connect SharePoint, use Gamma for presentations
  • RIGHT NOW: Use Act-Explain-Please for every AI interaction
  • WHEN YOU CAN: Install Claude and try Cowork for a month
  • WHEN YOU CAN: Set up one workflow automation
  • IN THE FUTURE: Play with AI tools like your career depends on it

Quick Wins You Can Do Today

  • Set up Custom Instructions in ChatGPT for your role
  • Turn off data training in all your AI tools
  • Create a prompt template for your most common task
  • Try one new AI tool you learned about today
  • Share Act-Explain-Please with one colleague
  • Upload one work document and ask AI questions about it
New

Google Stitch

Just announced — let's explore it together.

Google Stitch

  • Brand new — announced recently
  • This is the nature of AI — new tools launch constantly
  • Being comfortable exploring unfamiliar tools is the real skill
  • The frameworks we learned today work everywhere
  • Let's dive in and see what it can do...
Activity

Explore Google Stitch

15 min

Let's explore this new tool together as a group. What can it do? How does it compare to tools we already know? What use cases can you imagine?

Update

Midjourney v8

The latest version just dropped.

Midjourney v8

  • Significant quality improvement over v7
  • Better understanding of complex prompts
  • More photorealistic results
  • Improved consistency across generations
  • Note: Representational harms are still present
  • Always review AI-generated imagery for bias
Share

Teaching AI to Others

You now know enough to get others started.

The Five-Minute Pitch

  • You have 5 minutes to get a colleague interested. What do you say?
  • Start with Act-Explain-Please — simple and immediately useful
  • Explain: it's not fancy Google — it's an intelligence
  • You need to prompt it that way — give context and direction
  • Show them something YOU created with AI today
  • Demonstration beats argument every time

Dealing with Resistance

  • AI causes many people to feel frustrated and angry
  • Be cooperative — acknowledge frustrations
  • Then show something you've created — let results speak
  • Don't preach — meet people where they are
  • Some people will never be interested, and that's okay

Common Objections: Water & Copyright

  • Water: Data centres recycle most water, minimal vs agriculture
  • The real environmental issue is electricity, not water
  • Copyright: Synthetic data replacing human training data
  • It's becoming less of an issue with each model generation
  • It's unlikely you'll change anyone's mind on copyright
  • Acknowledge concerns, redirect to practical applications
Activity

Practice Teaching

10 min

Pair up. You have 5 minutes each to 'teach' your partner one thing you learned today. Use your laptop to demonstrate — don't just tell, show. Swap after 5 minutes.

Growth

Keeping Up

AI is changing constantly. How do you stay current?

Gemini 3.1, ChatGPT 5.5
& Nano Banana Pro

All launched in the past few weeks alone

Find Your People

  • Have fun with AI tools and learn from interesting people
  • YouTube: Matt Wolfe, Matthew Berman, Wes Roth, David Shapiro
  • Find LinkedIn people who share practical AI applications
  • Follow X feeds for your favourite tools
  • Join communities where people share what they're building
  • The best learning happens by experimenting and sharing
"

Have fun, push everything, and experiment.

There is currently no indication that AI progress is slowing down.

The pace of change will only accelerate from here

Homework

Homework

Things to do after today's workshop.

Your Action Items

  • Set up Act-Explain-Please as your default framework
  • Turn off training data in all AI tools (Settings → Privacy)
  • Try one new tool from today in your actual work this week
  • Create one workflow that automates a repetitive task
  • Share one thing you learned with a colleague
  • Follow 3 AI creators on YouTube or LinkedIn
  • Visit www.rupertchesman.com for resources

Resources

  • Course materials: www.rupertchesman.com
  • Vibe coding: www.rupertchesman.com/masteringai/vibecode
  • Google AI Studio: aistudio.google.com
  • NotebookLM: notebooklm.google.com
  • Gamma: gamma.app
  • Suno: suno.com
  • Midjourney: midjourney.com
Q&A

Questions

What would you like to know more about?

Thank You

Thank You

www.rupertchesman.com • © Rupert Chesman 2026