Trudy Graham

On New Year's Day this year, I decided I wanted to start this podcast. I'd built one before, but that was five years ago, and I didn't have time for a course or trial and error.

So instead of Googling "how to start a podcast" and drowning in generic advice, I gave ChatGPT a high-profile expert in podcasting and pointed it to their publicly available material. Then I asked it to build me a 30-day action plan based on what that expert had put out into the world.

What came back was specific episodes to listen to, blogs to read, activities mapped out day by day, built around what I already had and what I was missing. It felt like a mini course, tailored to exactly what I was trying to build. What would have taken weeks of trial and error compressed into a structured month.

We launched in February.

That's AI acting as a trainer, not an editor. Hand it a skill you need to build, point it at a real expert's material, and let it build you a personalised plan. Thirty days might not suit your reality in a school. Ten weeks, one task a week, might.

Most leaders only ever use AI to write and edit things. It can do so much more than that.

I share four other roles your AI tool can play, data analyst, researcher, coach, and consultant, on this week's episode of the Trailblazing School Podcast: https://trailblazingschools.riverside.com/e/5-ways-to-use-ai-as-a-school-leader-beyond-writing-and-editing

I built a business that, from the outside, looks like it's entirely about AI. I run workshops on it. I record podcast episodes and write newsletters about it, most weeks.

But that's not the thing that truly makes the impact.

A few months ago, I went back through everything I've built over the last three years. Every case study. Every testimonial. Every piece of feedback from the principals and middle leaders I've worked with. I was looking for the moment AI showed up as the thing that changed them.

It wasn't there. Not once.

What they talked about instead was getting out of the weeds and into the work that actually matters. Leadership teams stuck in meetings that were just information sharing, everyone reporting on their own silo, nobody actually thinking together. Small school principals carrying everything because there's no one else to hand it to. Not a personal failing, the research has been calling this a structural problem in school leadership for fifteen years.

AI didn't fix that for any of them. What fixed it was candour, structure, and a leader willing to let go of the busyness that felt urgent so there was room for the busyness that actually mattered.

It was a strange thing to discover about my own business. I thought I was selling a tool, but what I've actually been building, apparently, is space for people to think.

I get into what that looks like in practice on this week's episode: https://trailblazingschools.riverside.com/e/getting-out-of-the-weeds-and-into-the-work-that-actually-matters

Lisa reconnected with me in 2024. She had a newly shaped leadership team at her growing school in Walloon, and she told me straight: "I'm bogged down in operational, and that's not where I like to sit. I'm a big picture person."

Her team wasn't short on effort. They were, by her own account, amazing people who worked hard. What they were short on was candour. Their leadership meetings had become what she called "operational regurgitation": everyone reporting what they'd done, nobody challenging anybody else's thinking, everyone staying in their lane and calling it collaboration.

The first shift wasn't a tool. It was a distinction, the difference between teamwork and what I call "teamship". We rebuilt the meetings around tools that gamified challenge, and a framework to keep the team deliberate about which part of the work they were actually in.

A year later, Lisa calls it her most successful year in school improvement. She told her team directly: "I'm not interested in mini-me's. I need my thinking challenged." When she asked them afterwards whether they felt safe pushing back on her, they said yes, because she'd made it clear that was the job.

Yes, AI was part of Lisa's year. She uses it constantly now. But when I asked what actually changed her as a leader, she started with the team, not the tool.

Full story is on this week's episode: https://trailblazingschools.riverside.com/e/getting-out-of-the-weeds-and-into-the-work-that-actually-matters

Last year I found out something uncomfortable about my own leadership. Not from a colleague, but from ChatGPT.

I'd been using it for months to draft emails, find themes in sessions, work through tricky feedback. One day I asked it something different: to analyse all our conversations and tell me my strengths, my progress, and the gaps I might not be aware of.

The strengths section felt like an affirmation. Clarity of vision. Strategic intent. Audience sensitivity. Nice to read, but what I expected, really

The gaps section is what made me start to feel a bit uncomfortable. Over-layering. Hidden perfectionism in iteration. Emotional load management. Three things I recognised instantly and would never have said out loud, because nobody else has been close enough to my daily decisions to see the pattern.

A colleague softens things. A supervisor picks their moments. AI has no stake in managing my feelings. It just reports what's there, which is exactly why the gaps section was the most honest feedback I've had in years.

We've mostly used AI to move faster. This is different. It's not making me more efficient at who I already am. It's making me more aware of who I'm becoming.

If fifteen minutes with a prompt can show you a leadership gap nobody else was positioned to see, what else might it show you if you started asking better questions?

I share the exact prompt, and what I did with what it told me, on this week's episode. https://trailblazingschools.riverside.com/e/the-leadership-coach-you-already-have

Most leaders I work with are drowning in feedback about their school and starving for feedback about themselves.

Parent surveys. Staff wellbeing data. Review panel notes. Mountains of information about how the school is going. Almost none of it is about how *you* are going.

What you do get arrives twice a year, filtered through a supervisor managing thirty other leaders, softened enough that the real pattern gets lost.

So I've got a trick for you to try. If you've been talking to AI regularly, it's been paying attention. You've just never asked it what it noticed.

I ran an experiment on myself for six months. I asked ChatGPT to analyse all my engagements with it over the last 30 days: my strengths, my progress, and the gaps I might not be aware of.

The strengths section was affirming. The progress section was a list of small wins I'd taken for granted. The gaps section was honest, and it stung. (Over-layering. Hidden perfectionism in iteration. Emotional load management.) All of it true in a way nobody else has the vantage point to tell me.

This isn't AI telling you who you are. It's AI showing you your own paper trail, which you wrote.

➡️ Try it this week.
Open whatever AI tool you already use and paste in this exact prompt:
"Analysing all my engagements with you over the last thirty days, what have been my strengths? Where have I made progress? And what are the gaps I may not be aware of?"

Read the whole thing before you react to any one line. Reply in the comments if you try it, I'd love to know what it tells you.

School leader friends who are passionate about helping kids learn to read, consider completing this survey, share it in your network, and generally work your magic. Thanks, T.
______________________

Calling All Queensland School Leaders! – Share Your Insights on the Provision of Reading Interventions at Your School

QUT Ethics Approval Number: 7688

If you are a Principal, Deputy Principal, Head of Curriculum or other learning support leader currently working in a Queensland Primary School and have insights into how your school supports students with persistent reading difficulties, you are warmly invited to participate in this study.

This research is being conducted by Cheryl Hooper, a doctoral candidate in education, as part of her studies toward the completion of the degree of Doctor of Education.

The purpose of this research project is to explore how schools are implementing interventions for students with persistent reading difficulties.

The survey takes around 30 minutes to complete and includes choosing from set responses with opportunities for extended responses. Responses are confidential and you will be deidentified in any reports or journal articles.

Click here to complete the survey: https://qsurvey.qut.edu.au/jfe/form/SV_b8ip47LLjVnvrH8

To all my colleagues, thank you for the amazing work you do, the difference you make and the impact you lead.

Last year I went down a rabbit hole naming my AI tool. I wanted a name that would remind me, every single time I opened it, that I was talking to something built on uneven, incomplete, human-shaped data.

I landed on Chad.

ChatGPT itself described Chad as helpful, articulate, occasionally overconfident, trained on a mountain of internet text that skews Western, English-speaking, and male. Chad loves a trend and a buzzword and a neat-sounding answer. He can sound completely certain while being biased, inaccurate, or incomplete. A fast talker, not always a deep thinker.

Naming him that way wasn't a joke at the tool's expense. It was a habit. The name does some of the critical thinking for me before I've even typed a prompt.

Here's the challenge for this week. Open whatever AI tool you use and ask it directly: what are the biases in your training data? Not a vague sense that something might be off, an actual, specific list.

Reply in the comments with what your AI admits to.

In the early nineties, I was a teaching principal in Burketown, a remote school in Queensland's Gulf of Carpentaria. Thirty-three students, mostly Aboriginal and Torres Strait Islander kids.

I went looking for books where they could see themselves on the page. Their family structures, their version of an ordinary week: fishing, camping, swimming, groceries arriving once a week on the back of an unrefrigerated truck from a town five hours away. I searched the library. I searched publisher catalogues. I searched conference stands. I found one set of readers that came close.

Fast-forward to now and there are far more choices, far more representation.

That gap didn't close because publishing got less biased on its own. It closed because someone went looking for what was missing and built it.

I think about that constantly with AI. Every large language model learns from whatever fraction of human-written text got recorded and kept, not from reality itself. That's not a glitch sitting in the code somewhere. It's baked into the training data, because the data is us, the best and the worst of what's been written down.

The same gap I found in a library in the nineties exists in every AI tool you use today. It's just built out of data instead of a publisher's catalogue.

I unpack what I found on this week's episode. https://trailblazingschools.riverside.com/e/uncovering-bias-in-ai

Dr Joe Azzopardi told me about a young adult group he used to help run at church. No one over 30, which also meant no mentors. When it came time to share a meal together, people would bring chips and cookies, because that's what you can afford when you're just starting out and nobody older is contributing.

Whenever someone in the group had kids, they'd leave and join another community, because there was nothing there for families. The group had plenty in common, but it had no range. Joe realised the problem wasn't effort. It was that everyone was at the exact same life stage.

So they made it multigenerational. And almost immediately, things started to work. There were people with real incomes to help out. People willing to say "let me teach you that." Kids running around, with people willing to watch them.

I got me thinking about how closely that maps onto the regional Queensland schools I work with, staffed almost entirely by early-career teachers because that's who can move to a small town with no mortgage and no family to relocate. No one to mentor them. We already know that's one of the biggest factors in losing young teachers in their first five years.

Joe's research digs into exactly this, why generational range matters more than we give it credit for. Full conversation is on this week's episode:https://trailblazingschools.riverside.com/e/leading-across-generations-with-dr-joe-azzopardi