Practical AI Masterclass Playback: What We Covered + Watch the Recording

On Wednesday 30 April 2026, I ran a free Practical AI Masterclass for a small group of brilliant humans — business owners, consultants, and leaders who are genuinely curious about AI but wanted something more than a tool tutorial. They wanted to understand how it works, why it works, and — most importantly — how to use it without losing themselves in the process.

The recording is below. And because I know not everyone has an hour to watch right now, I've written up the key ideas we covered so you can dip in and out.

Watch the Recording

Full recording — 58 minutes.

What We Covered

1. The Sunset Exercise: Why Your Lived Experience Is Your Greatest AI Asset

We opened with something unexpected. I asked everyone to close their eyes and picture a sunset — and then tell me what they saw.

Emma described warmth and the feeling of rays on her skin. Gillian saw the sea and "wonder." Tara (calling in from the West Coast) described pinks and purples as the clouds caught the last light. Victoria painted a vivid Thai beach scene with pastel gradients and an almost-red sun meeting the ocean.

Why does this matter for AI? Because when you describe a sunset, you don't just recite facts — you draw from your own lived experience, your own memory, your own felt sense. And that is exactly what AI cannot do.

I also shared something personal: I have a condition called aphantasia, which means I don't see mental images. When you say "picture a sunset," I don't get a picture. I get information — memories, associations, descriptions. In a lot of ways, I process the world more like an AI does. And understanding that has completely changed how I teach people to use these tools.

2. How AI Actually Thinks: Vectors, Neighbourhoods, and Tokens

This is the bit that made Gillian's brain click. I explained that AI doesn't think in pictures or sentences — it thinks in vectors.

When you type a word into a large language model, it converts that word into a mathematical vector and sends it to a "neighbourhood" — a cluster of related concepts, associations, and meanings. The AI then looks at what exists in that neighbourhood and constructs the most statistically probable response.

This is why the way you craft your prompt changes everything. If you type vague, generic words, the AI goes to a generic neighbourhood. If you type specific, rich, personal language — your actual words, your actual experience — it goes somewhere much more interesting.

Tokens are the economy of AI. They're the units of processing that power everything you do with these tools. They matter more once you start building automations and agents, but even in everyday use, understanding tokens helps you understand why paid tools behave differently from free ones — and why memory, context, and consistency improve the more you invest in your setup.

3. Your Human User Manual

One of the most practical things I shared was the concept of the Human User Manual — a document I've built for myself that I upload into any new AI tool I start working with.

It includes who I am and my life story, my preferences, dislikes, and working style, examples of my best work (real examples, not descriptions), my brand voice, tone, and creative direction, and how I like to communicate and what "great" looks like to me.

The reason this works so well is that AI is deeply programmed to please you — but it can only please you if it knows what pleasing you actually looks like. Without this context, it's guessing. With it, it's extending your genius.

I'll be sending the Human User Manual template to everyone who attended the masterclass. If you didn't attend but want a copy, take the Wonder Audit — it's the first step in building yours.

4. Wonder Conducting: The Philosophy Behind the Method

I don't call what I teach "how to use AI like a boss babe" (though you do learn that). I call it Wonder Conducting because the goal is to extend your reach, not replace yourself.

The distinction matters. AI has limitless access to everything that has ever been digitised. But the only thing that differentiates how you use it from how I use it is your point of view, your taste, your creative direction, your vision, and what you can imagine wanting to create.

That's why I say the greatest currency in a world of AI is taste and creative direction. The tools are equalised. The humans aren't.

We also talked about what AI genuinely cannot do: it cannot feel. It doesn't know what it's like to break up with someone, to sit with an elder from the Anangu tribe and receive only level-one knowledge, or to feel the specific coolness of ocean spray mixed with afternoon sun. Only you know that. And that felt sense — when you bring it deliberately into your prompts — creates something AI-generated content almost never achieves: meaning.

5. Hallucinations, Temperature Settings, and Not All LLMs Are Equal

We got into the mechanics of why AI sometimes confidently tells you something completely wrong. I shared my own experience of an AI insisting — even when I pasted in news articles — that a public figure was still alive when they weren't. That's hallucination: the model using probability to give you the most likely next answer, even when that answer is factually incorrect.

Different tools have different temperature settings — essentially, how creative and how risky the model is willing to be. Grok runs hot (great for creative sparring, accesses Twitter in real time). Claude is more measured. ChatGPT sits somewhere in between. Knowing this helps you choose the right tool for the right job rather than assuming they're all the same.

6. Agentic AI: The Grad on Acid Steroids

The session's most memorable metaphor: an AI agent is like hiring the most enthusiastic grad you've ever met, giving them access to everything, and then forgetting to give them a job description.

They will go off and do something. It just might not be what you wanted.

I've been brought in to clean up after three clients who spent over $100,000 implementing autonomous AI without first mapping their processes, their customer journey, or their SOPs. The technology isn't the problem — the lack of human clarity before deployment is.

My recommendation: start with agentic AI inside a structured tool like Notion or ClickUp, where you've already defined what the work is. Run experiments. Understand what you're asking it to do. Then scale.

I shared a real example: one of my Notion agents sent me an alert last week flagging that a client had requested out-of-scope work, estimated the extra hours required, and drafted an email to the client asking for approval — all without me asking it to. In the past, I would have either done the work for free or not done it at all. That's the power of a well-briefed agent.

7. The Traffic Light System for AI Risk

Green, orange, red.

Green — experiment freely. Welcome emails, website chatbots, content drafts, research summaries.

Orange — proceed with awareness. Customer-facing communications, anything involving personal data, integrations that create "doorways" between tools.

Red — always keep a human in the loop. Anything involving sensitive client information, legal or financial decisions, or communications where a mistake would damage a relationship.

I also talked about the concept of doorways: every tool you connect to your AI stack is a doorway into your digital house. You don't need to be paranoid — but you do need to be aware of which rooms you're giving access to, and which ones you're keeping locked.

8. Your Homework

Before you open another AI tool, do this:

Connect with a career success moment. Something you're genuinely proud of. What made it great? The people, the process, the creative freedom, the constraints? These are your vectors.

Do the sunset exercise with someone. Over coffee, in a meeting, wherever. Ask them to describe a sunset. Listen for the specificity, the lived experience, the felt sense. That's what you're bringing to AI.

Find one process in your life that could be one version better. Not a full automation overhaul — just one thing. A spreadsheet that could be an app. A folder system that could be a dashboard. Ask AI: how could I make this one version better?

Start your Human User Manual. Who are you? What does great look like for you? What are your preferences, your non-negotiables, your working style? The more you put in, the more you get back.

Want to Go Deeper?

If this session sparked something for you, there are two obvious next steps:

Take the Wonder Audit — a free 10-minute assessment that maps your current AI confidence, identifies your biggest gaps, and gives you a personalised starting point. It's the first step toward building your Human User Manual.

Take the Wonder Audit →

Join Wonder Conductor — my 4-week AI mastery sprint for conscious leaders and business owners. We cover everything from your Human User Manual and AI stack to agentic workflows, automation, and building real tools with AI. The next cohort starts Tuesday 6 May 2026.

Explore Wonder Conductor →

Sarah Pirie-Nally is an AI strategist, keynote speaker, and the founder of Wonder & Wander. She helps leaders use AI to extend their genius — without losing themselves in the process.