I built an AI-powered school for my daughter, and somewhere between the prompts, the dashboards, and the sleepless nights over her education.I realised the problem wasn’t artificial intelligence.
It was artificial education.
The Breaking Point
My daughter left school last year. Not for some enlightened gap-year experiment, but because of something painfully ordinary: bullying.
The school’s response was exactly what you’d expect from a 19th-century institution wrapped in 21st-century PR.
A lot of polite concern and zero meaningful action.
They fined me for taking my kids out of term time for holidays, but didn’t blink when the same system drove her out entirely.
So, out of frustration, necessity, and probably mild insanity, I built her an AI-powered school. Not because I wanted to “disrupt education”. I just wanted her to learn in peace.
What started as a workaround turned into a revelation.
The Education Illusion: Knowledge Isn’t Learning
Once I began mapping the national curriculum, I had a horrifying realisation: the entire system is designed for information transfer, not intelligence development.
Every subject, every exam, every lesson, all built on the same logic. Here’s a body of known knowledge, memorise it, repeat it, move on.
It’s an assembly line. You’re not building thinkers, you’re manufacturing parrots.
If schools were a brand, they’d be the worst kind, one that still advertises “innovation” while selling the same tired product in a shinier box.
The whole construct was built for the Victorian era: sit down, shut up, absorb, repeat.
It made sense when education was training factory workers. But in 2025, the “fact factory” model is insane.
AI can now access, cross-reference, and explain every GCSE topic ever written in milliseconds, yet we still force kids to memorise it like monks transcribing the Bible.
Education hasn’t just fallen behind. It’s forgotten what it’s for.
AI Was the Easy Part
When I started building the AI system, I thought the hard bit would be the tech. Wrong. That part was relatively simple.
Free educational content is everywhere. I could plug in data, design custom GPTs, even generate teaching assistants that could answer questions, set tasks, and provide instant feedback.
Within weeks, my daughter had a functioning digital school: timetable, lesson structure, assignments, the lot.
But it was soulless. Efficient, yes. But lifeless.
The first prototype taught like a robot because, of course, it was one. It could deliver perfect answers, but not meaningful ones. There was no arc to the learning, no continuity, no connection.
Lessons existed in isolation, and the system kept forgetting what had come before.
It wasn’t a school. It was a content vending machine.
So I had to fix it. Not by adding more AI, but by giving it a spine. Literally.
The “Spine” and the Problem With AI Product Design
Every GPT I built was good at its specific job but terrible at remembering context. Each one acted like a teacher who walks into class every morning with amnesia.
So I built what I called the spine. A central, data-driven backbone that every GPT could reference.
The spine carried the timetable, lesson history, and progress tracking for every subject. The system finally had structure. Lessons connected. Students could progress through logical stages.
That was my first major lesson in AI product design. If you’re not obsessed with context, you’re just building smarter chaos.
But structure wasn’t enough. When my daughter began using it, her feedback was blunt: “It’s boring.”
Fair point.
So I iterated again. I built a prompt-design tool that structured every teaching interaction around scaffolding, starting with support, then gradually removing it as she learned.
Instead of dumping information, the AI became a coach. It asked questions, encouraged, and adjusted tone.
And something shifted.
She started engaging. Asking questions. Getting curious again.
That’s when I realised the goal wasn’t to make the AI more intelligent. It was to make the learning more human.
When Real People Use Your Product, It Gets Real Fast
Then something unexpected happened. Another parent, whose daughter had also been bullied out of another school, asked if she could join.
Now I had two students, two radically different users.
One was previously disengaged with real school and was more worried about where the next verbal attack would come from, and one was gifted academically but not able to focus due to abuse.
One needed emotional safety, the other needed challenge.
Instantly, my system broke. The AI couldn’t balance different learning styles, emotional cues, or pace. It wasn’t failing technically. It was failing empathetically.
That’s when I learned my most painful product truth: AI doesn’t fail because it’s dumb. It fails because it doesn’t care.
I had to rebuild again, integrating behavioural data, motivational design, and daily feedback loops. I added a dashboard, playlists, motivational quotes, even physical activities. Slowly, it became something more than digital schooling. It became a holistic environment.
In that moment, I stopped being an engineer and became a strategist.
Market Orientation for Machines
Everything that worked came from listening, not coding.
I stopped trying to force the system to behave like a teacher and instead started asking, ‘What does the learner actually need?’
Schools don’t do this. They’re designed for regulators, not students. That’s why they fail.
AI, ironically, provided me with the data that schools often fail to collect: performance, engagement, mood, and progression.
I didn’t need a parents’ evening because I had constant insight. When she struggled, I knew where. When she thrived, I knew why.
And that’s when it hit me. Education isn’t broken because it lacks technology. It’s broken because it lacks curiosity.
Next Step: From Replication to Creation
I’m now evolving the platform into something much bigger. Not an AI version of school, that’s the mistake everyone makes. But a replacement for what school should be: a system designed for human success, not bureaucratic compliance.
The next phase is about teaching creation, not replication.
We’re integrating entrepreneurial modules, financial literacy, interpersonal skills, and even tools for managing anxiety and social dynamics, all built around the central idea of self-directed learning.
AI won’t just be a tutor. It’ll be a builder. A designer of bespoke curricula for every learner based on their strengths, goals, and progress.
Imagine an education system that dynamically adjusts to each child, not the other way around. That’s the vision.
And yes, it’ll take time.
Guardrails, UX, testing, safety nets, all the boring bits that make something real. But after what I’ve seen, I’m convinced this is where education will end up. The only question is who gets there first.
The Human Lesson
People ask if I’m worried that my daughter is learning from AI. Not really.
I see every lesson she does, every mistake, every success. I know more about her education now than any parent who gets a five-minute report card once a year.
She’s happier. Fitter (she uses AI to build her own fitness workouts). More curious. She’s back to learning because she wants to, not because someone with a clipboard said she must.
And me?
I learned that building an AI school isn’t about teaching machines to think. It’s about remembering how humans do.
The future of education won’t belong to those who teach machines to talk. It’ll belong to those who teach humans to think again.
Thanks for reading.
Andrew Holland
Side note:
I keep getting asked if I’m interested in turning this into a product.
The reality is that education is a complex business and complex sector. I’ve learned a lot doing this; I’m still learning, and I do this in my spare time.
So, while I do believe there is something solid here and a product could be created that helps children both in and out of schools, and a revolution in the education sector could happen, I think schools themselves are too far behind to fully embrace this.
Only if the education secretary themselves gave me a call and asked me to come in and discuss what we’d learned, then I would help and show this.
The reality is that I built something suitable now for a couple of people. The idea of creating a product where parents would pay to take on the responsibility of educating hundreds of students would be too much.
The thing is that this has taught me a great deal, but am I ready to turn it into a paid product?
Certainly not. That would require significant financial investment, education specialists, and, indeed, a comprehensive review of education itself from a top-level perspective.
For now, however, it serves my purposes; it’s helping my daughter, and yes, I’ve replaced school in a way that I don’t think schools have figured out yet, and I think she’s going to benefit from this far more than she would from a full-time traditional school education.
Originally published on LinkedIn on 8 October 2025.