Sarah Pirie-Nally
Back to Blog
20 April 2026 12 min read

How to Conduct AI Like a Pro

A live masterclass breakdown: 6 lessons on AI workflows, Custom GPTs, tool chaining, and the human edge that no algorithm can replicate.

How to Conduct AI Like a Pro
Sarah Pirie-Nally

Sarah Pirie-Nally

AI Strategist Β· Keynote Speaker Β· Author

What if you could compress weeks of professional work into a single afternoon? Not by cutting corners β€” but by learning to conduct AI the way a composer conducts an orchestra. That's exactly what happened inside this live AI Masterclass session, and I'm going to take you through every lesson, every tool, and every mindset shift that made it possible.

Watch the full session above, or read the masterclass breakdown below.


Before We Begin: The One Principle That Changes Everything

Every tool I'm about to show you is secondary to this single idea: AI should unlock your humanness, not replace it.

The people who are winning with AI right now are not the ones who have handed over their thinking to a machine. They are the ones who have learned to use AI to handle the routine β€” the research, the first drafts, the formatting, the synthesis β€” so they can reinvest that freed time into the work that only a human can do. The relationship-building. The creative leaps. The judgment calls that require lived experience and emotional intelligence.

"AI is a productivity multiplier β€” but only if you know what you're multiplying. If you multiply mediocrity, you get faster mediocrity. If you multiply your best thinking, you get something extraordinary."

Keep that in mind as we move through each lesson. The tools are powerful. But you are the conductor.


Lesson 01

The "Zero to One" Principle β€” From Idea to Prototype in Minutes

What happened in the session: To open the masterclass, I wanted to shatter the audience's expectations of what's possible. I asked CT β€” founder of The Moore Centre, a clinical psychology practice β€” to describe her business in a single sentence. I pasted it into Relume β€” and within 90 seconds, we had a complete website: sitemap, wireframes, styled design, ready to export to Figma or Webflow.

The room went quiet. That silence is always the moment I know something has landed.

The lesson underneath the demo: The distance between a thought and a tangible thing has collapsed. What used to take a design team six weeks now takes a single afternoon. But here's what most people miss when they see a demo like this: the quality of the output is entirely determined by the quality of the input.

πŸ”‘ Key Insight

Garbage in, garbage out β€” but the inverse is also true. A specific, well-crafted prompt is the difference between a generic template and a genuinely useful prototype. Prompting is a skill. It is, as I said in the session, like "painting with words." The more vivid and precise your description, the more useful the output.

Your homework from this lesson: Pick one thing you've been procrastinating on because it felt too big to start β€” a website, a proposal, a course outline, a business plan. Write a single, specific paragraph describing what it is and who it's for. Then paste it into an AI tool and see what comes back. You don't have to use the output. But you'll never look at a blank page the same way again.


Lesson 02

Custom GPTs β€” Building Your Specialised AI Team

What happened in the session: This was the centrepiece of the masterclass β€” and the part that generated the most "I had no idea you could do that" moments. Standard ChatGPT is a generalist. It knows a lot about everything and not enough about you, your business, your voice, or your specific context. Custom GPTs change that entirely.

I walked the group through building a Custom GPT from scratch: uploading a PDF knowledge base, writing specific behavioural instructions, defining the tone, and β€” critically β€” turning off web search so the AI only draws from your uploaded documents. The result is a closed-loop assistant that thinks and speaks in your world, not the internet's.

I showed them two of my own: a Brave Leader Coach built from a specific coaching framework, and β€” my personal favourite β€” a Custom GPT I built as an operating system for my daughters. Not a chatbot. An operating system. One that encodes our family's values, our communication style, our decision-making frameworks. That's what happens when you stop thinking of Custom GPTs as tools and start thinking of them as systems.

"Standard ChatGPT is like hiring a brilliant generalist who knows nothing about your company. A Custom GPT is like that same person after six months of deep onboarding. Same intelligence. Completely different usefulness."
πŸ”‘ Key Insight

Give your AI a backbone. The single most common mistake people make with AI is treating it like a search engine β€” asking it questions and hoping for the best. The people getting extraordinary results are the ones who invest time upfront in defining exactly how their AI should behave: what it should reference, what it should never say, what tone it should use, and what constraints it should operate within. That investment pays back every single time you use it.

The step-by-step process I demonstrated:

  1. Navigate to ChatGPT β†’ Explore GPTs β†’ Create
  2. Write a clear system prompt: who this GPT is, what it does, what it should never do
  3. Upload your knowledge base (PDFs, documents, frameworks, past work)
  4. Turn off web search if you want a closed-loop system
  5. Define the tone: warm, direct, humorous, formal β€” whatever fits your context
  6. Test it. Break it. Refine it. Repeat.

Your homework from this lesson: Design your own Custom GPT. It doesn't have to be perfect. It doesn't have to be complex. Start with one specific use case β€” a writing assistant that sounds like you, a research tool for your industry, a coach built around a framework you love. Build it, use it for a week, and notice what shifts.


Lesson 03

Human-in-the-Loop Design β€” Why You Should Never Fully Trust Your AI

What happened in the session: CT asked a question that I think every thoughtful person asks when they start working seriously with AI: "How do you make sure the output is actually good?"

My answer was the Bias Checker β€” a Custom GPT I built specifically to review AI-generated research for Western and biomedical biases before it goes anywhere near a client. This is what I call meta-AI: using AI to quality-check AI. It sounds recursive, but it's one of the most sophisticated things you can do with these tools.

The deeper principle here is what I call Human-in-the-Loop design. You are not abdicating your judgment to the machine. You are designing a system where the machine does the heavy lifting and you apply your expertise at the critical checkpoints. The AI drafts; you direct. The AI researches; you evaluate. The AI formats; you decide what's worth keeping.

πŸ”‘ Key Insight

Build your own quality control layer. Before any AI output leaves your hands, it should pass through a human checkpoint β€” or a specialised AI checkpoint you've designed for that purpose. Ask: Does this reflect my actual values? Is this culturally appropriate for my audience? Does it sound like me? The speed of AI is only an advantage if the output is worth sending.

The Bias Checker GPT I demonstrated was built around a specific brief: catch assumptions embedded in research that reflect a narrow Western, biomedical worldview. For the work I do with neurodiversity, this matters enormously. But the principle applies everywhere. You can build a Bias Checker for gender assumptions, for cultural context, for industry jargon, for brand voice. The constraint is only your imagination.


Lesson 04

Tool Chaining β€” Where the Real Magic Happens

What happened in the session: This is the lesson that separates people who use AI from people who are genuinely transformed by it. Most people use AI tools in isolation. They ask ChatGPT a question, get an answer, and move on. What I demonstrated in this session was something different: a complete workflow where each tool's output feeds directly into the next.

Here's the pipeline I showed live β€” using research for CT's clinical work as the real-world brief:

πŸ’¬
Specific Prompt

A precise research question about OCD and tics in children

β†’
πŸ”
Grok (xAI)

Deep academic research β€” peer-reviewed papers, citations, nuanced findings

β†’
βœ…
Bias Checker GPT

Quality control β€” checking for cultural and biomedical assumptions

β†’
🎨
Gamma.app

Instant professional presentation β€” structured, designed, ready to share

β†’
✨
Professional Deliverable

A research-backed, bias-checked, beautifully designed presentation β€” in under 20 minutes

That pipeline would have taken a junior researcher two full days. We did it in a live session, in front of a room of people, in real time.

"No single AI tool does everything well. The sophistication is in knowing which tool excels at which task β€” and designing a workflow that plays to each tool's strengths."
πŸ”‘ Key Insight

Think in workflows, not tools. The question is not "which AI tool should I use?" The question is "what is the workflow I'm trying to create, and which tools are best suited to each stage of it?" Map your current work processes. Identify the stages that are most time-consuming and least dependent on your unique human judgment. Those are your automation candidates.


Lesson 05

Relational AI β€” Onboarding Your AI Like a New Team Member

What happened in the session: Towards the end of the masterclass, I made a point that I think is the most underrated insight in the entire field of AI productivity: the quality of your relationship with your AI determines the quality of its output.

Working with an untrained AI is like an awkward first date. It doesn't know you. It doesn't know your context, your voice, your values, or your goals. It's doing its best with almost no information. The people who get extraordinary results from AI are the ones who invest time in onboarding it β€” just as you would a brilliant new team member.

That means creating a Language Guide: a document that captures your tone of voice, your most-used phrases, examples of your best writing, your brand values, your audience. Upload it. Reference it. Make it the foundation of every Custom GPT you build. The AI will start to sound like you β€” because it's been trained on you.

πŸ”‘ Key Insight

Create your Language Guide this week. Open a document and write: your communication style in three words, five phrases you use constantly, three things you would never say, your audience and what they need from you, and two examples of your best writing. Upload this to every AI tool you use regularly. The difference in output quality will be immediate and significant.


Lesson 06

The No-Code Frontier β€” Building Apps Without Writing a Line of Code

What happened in the session: I closed the masterclass with a glimpse of what's coming next β€” and what's already here for those willing to explore it. Tools like Lovable and Replit now allow you to build functional software applications using nothing but natural language. I showed a NeuroSparkles quiz app β€” a real, working application β€” built without a single line of code.

This is not a future possibility. It is happening now. And it means that the barrier between "I have an idea for a tool" and "I have a tool" has essentially collapsed.

"The question is no longer whether you can build something. The question is whether you can clearly describe what you want to build. That's a thinking skill. That's a human skill. That's your skill."
πŸ”‘ Key Insight

The bottleneck has moved from technical skill to creative clarity. You no longer need to know how to code to build software. You need to know how to think clearly about what you want, who it's for, and what it should do. That clarity β€” that ability to articulate a vision precisely enough for a machine to execute it β€” is the new technical skill. And it's one you already have the foundation for.


The Real Takeaway: Systems Thinking, Not Tool Collecting

If you've been reading this as a list of tools to try, I want to gently redirect you. The tools will change. They are changing right now, faster than any of us can track. What won't change is the underlying principle that makes all of this work:

You are the conductor. The AI is the orchestra.

The most valuable thing you can develop right now is not proficiency with any single tool. It is the ability to think in systems β€” to look at your work and ask: what is the workflow here? Where is the human judgment irreplaceable? Where is the routine work that AI could handle? How do I design a pipeline that plays to the strengths of both?

That is the work of a Wonder Conductor. And it is learnable. I know, because I watch people learn it every single cohort.

Ready to build your own AI workflow?

Wonder Conductor is my 12-week program for leaders who want to go beyond the tools and build a genuine AI strategy β€” one that amplifies their humanness and transforms their professional output. The next cohort starts 5th May 2026.

Secure Your Spot β†’

Your Masterclass Action Plan

Don't let this be another article you read and forget. Here are the six actions β€” one from each lesson β€” that will move you from consuming AI content to actually using it:

  1. Write the paragraph. Describe one stalled project in specific, vivid detail. Paste it into an AI tool. See what comes back.
  2. Build your first Custom GPT. One use case. One knowledge base. One clear set of instructions. Use it for a week.
  3. Design a quality checkpoint. Before any AI output leaves your hands, what question will you ask to verify it meets your standard?
  4. Map one workflow. Choose a recurring work task. Draw the stages. Identify where AI could accelerate each one.
  5. Create your Language Guide. Tone, phrases, examples, values. Upload it everywhere.
  6. Try a no-code tool. Describe an app or tool you've always wished existed. Paste it into Lovable or Replit. See what happens.

The gap between where you are and where you want to be with AI is not a knowledge gap. It's an action gap. Close it β€” one experiment at a time.

_
Pip the AI mascot

Free Β· 10 Minutes

Discover Your AI Readiness Score

The Wonder Audit gives you a personalised score across 5 dimensions of AI leadership β€” so you know exactly where you stand and what to do next.

Take the Free Wonder Audit
Sarah Pirie-Nally

Sarah Pirie-Nally

AI Strategist Β· Keynote Speaker Β· Author Β· Founder, Wonder & Wander

Sarah helps leaders and organisations harness the power of AI without losing what makes them irreplaceable β€” their humanity. She has spoken on 6 continents, built the Wonder Conductor program, and runs fortnightly Practical AI masterclasses attended by 550+ leaders.

Enjoyed this article?

Get weekly insights
in your inbox

Join 550+ leaders getting Sarah’s weekly take on AI, human intelligence, and the future of work.