Sal Khan told TED years ago. AI as a personal tutor for everyone. One-on-one. Accessible. Research says it lifts grades.
This year? Silicon Valley nodded. San Diego nodded. Founders. Educators. Investors. All agreed. AI raises the floor. No more bad teaching at scale.
I mostly agree.
The data isn’t thin. Recent trials show effect sizes. Meaningful ones. The optimism isn’t just hype, it’s grounded in reality. The tech people are pointing somewhere real.
But they are looking at the wrong variable.
I know this because I built live coding instruction at scale. Not theory. Live sessions. I saw the AI tutors. The lesson plan generators. The disengagement signals. The automated progress reports. I watched what they did. I watched what they didn’t.
My takeaway?
AI’s biggest win in education isn’t for the student. It’s for the teacher.
What AI Actually Does
Let’s look at the tech.
AI generates content. Fast. It writes reports that used to eat up hours of a teacher’s evening. It sends metrics to parents. It routes kids to practice problems based on how they failed last Tuesday.
I’ve built all of these. I’m not criticizing from a hilltop.
They work. As infrastructure.
But here is the cold hard truth I learned: these tools improve the system around the teaching. They don’t improve the teaching. They make the machine run faster. The fundamental exchange between a child and a mentor? That stays the same.
The Chaos of Live Class
Scaling live instruction isn’t a coding problem.
It’s a human mess.
Software manages it, poorly. Always imperfectly.
Think about the coordination. Time zones. Academic calendars that never line up. Home environments. You’re trying to match a teacher with a student across continents.
An algorithm tries. It fails to capture the nuance. You need to match on teaching style. Pace. Personality. How deep do they know the subject?
Then there is the noise.
Connection drops. Not everywhere. Everywhere, just differently. Students get busy. Teachers get busy. School demands pile on top of platform demands. Home life changes on both sides of the screen simultaneously.
It’s exhausting. It’s relentless.
The Variable That Matters
Through all the chaos, one thing stayed consistent.
The teacher.
John Hattie analyzed 800 meta-analyzes. He found teacher effectiveness drives student achievement more than anything. More than class size. More than tech access. More than curriculum design.[1]
We ran millions of sessions. We saw the same thing. Teacher quality predicted outcomes. Everywhere. Every age. Every subject.
If the teacher is bad, the product is bad. Period.
Scaling Culture
When you have five teachers, you know them. You love them. You fix their issues.
When you have five thousand, you can’t.
What scales is culture. Or rather, the system that carries it. Onboarding. Feedback loops. Recognizing good work.
The teachers who win are the ones who own their classroom. The ones who get feedback that actually makes them sharper. The ones who have peers to talk to.
Building that trust took longer than building the code.
Fix the Teacher, Not the Student
So where do we put the AI?
Not in front of the kid.
In front of the teacher.
The best AI doesn’t chat with students. It saves teachers two hours of lesson prep. What does a teacher do with two free hours? Patience. Presence.
AI spots disengagement early. A signal the teacher wouldn’t catch for weeks.
AI writes the parent report in three seconds. The teacher spends those three hours talking to the parent instead of typing.
That compounds. That changes learning outcomes. Because it leverages the variable that actually matters.
The Centaur Model
There’s a term in chess: Centaur Play. Human and AI together. Stronger than either alone.
Edtech keeps playing tug-of-war. One side says AI replaces the tutor. The other says tech is just a digital textbook.
Both are wrong.
The future is the Centaur Teacher.
An AI tutor alone lacks the human nudge. It sees a pause; it doesn’t see frustration. It can’t share a cultural reference to motivate a kid who feels unseen.
A human teacher alone lacks infinite bandwidth. They can’t track 30 micro-progression lines. They forget what a child got wrong in March.
Put them together.
The teacher brings empathy. The social-emotional bridge. The reason a kid wants to learn.
The AI brings the eyes. Real-time data. Automated prep. The safety net.
This isn’t about how much AI you have.
It’s whether the AI makes the teacher better.
AI is the most powerful tool to democratize quality, but only as an amplifier. It doesn’t replace the human; it frees them to be seen.
We automate the process. We liberate the professional.
The best technology doesn’t make a student stare at a screen. It helps them look a teacher in the eye and feel known.
[1] Hattie, John (2009)
[2] RAND (2024)
[3] UNESCO (2024)
[4] Escueta et al (2017)
