Salman Khan’s first attempt at an education revolution was in 2006.

He was a hedge fund analyst but had learned the power of teaching digitally while tutoring his 12-year-old cousin in algebra. He was living in Boston and she was in New Orleans, so he tutored her via online messaging.

Soon more of his cousins were seeking his help, particularly with maths.

“They had great teachers but they were still struggling,” Khan told an event that was held in London in June this year to mark the publication of his new book, Brave New Words: how AI will revolutionise education (and why that’s a good thing). “What I saw over and over again were gaps in their knowledge, either things that they never learned well in the first place or things they had just forgotten.”

Once Khan personalised their learning, his cousins quickly improved. He began using software to write practice problems for his cousins to solve - and for him to monitor their progress. He then started recording videos of his lessons and uploading them to YouTube.

The Khan Academy and flipped learning

He quickly recognised that lots more students might benefit from this approach. And so he set up Khan Academy, a website offering maths videos.

It was founded as a non-profit organisation and within a year of Khan quitting his day job to run the platform full-time, the academy, which is based in Silicon Valley, California, had received significant grants from Google ($2 million, equivalent to £1.6 million) and the Bill and Melinda Gates Foundation ($1.5 million).

• Salman Khan: Teaching’s master of the universe
• Edtech: How primary schools can best use AI
• Webinar: Why school leaders need to understand what AI actually means

With Khan Academy, Khan popularised “flipped learning”, the idea being that students are given video lectures to watch at home before their lesson. Then, in the classroom, the teacher is freed up to facilitate applied learning, where students work on real-world problems.

Teachers wrote to Khan to tell him that his videos had enabled them to adopt this approach, he said in a TED Talk in 2011. “Now [the classroom] is a human experience,” he said, where teachers and students “are interacting”.

At that time Khan had high expectations about how Khan Academy and its associated new approaches to learning might revolutionise education, he tells Tes.

“Anyone who starts anything - non-profit, for profit - you have to have a bit of delusional grandeur,” he says. “I was like, ‘This could be huge. This could touch a lot of people.’”

But still the scope of Khan Academy, which now offers science and humanities classes as well as maths, has surpassed his expectations.

“If you told me, in 2007 or 2008, that Khan Academy would have 150 million registered users and videos in 50 plus languages, I would be like, ‘Those are some of my biggest dreams.’”

The numbers are significant. Khan estimates that “a majority” of maths teachers in the United States use Khan Academy. There are about 100,000 teachers working in the American school districts that use a paid-for version of the platform, which offers tailored programmes and technical support, he says.

“It’s become best practice that you should be doing active learning in the classroom”

But in reality Khan Academy has not transformed teaching like Khan hoped it might. In the US, as in the UK, students still typically sit at desks while a teacher delivers a lecture-style presentation, and then they complete tasks based on what they have learned.

“If you walk into a random classroom, for the most part it seems pretty similar to what we used to see,” he says. “If you asked me 10 years ago, I would have hoped… I mean, I’ve given TED Talks saying you shouldn’t need to give lectures any more, and everyone should be able to go at their own pace.”

While it may not be radical, he insists “there has been a change”, however. “I don’t want to say this is because of Khan Academy - we might have helped on the margin. But 10 years ago no one was really questioning the traditional lecture model, and no one was really talking about a lot of these gaps that students have in their learning.”

“Now it’s become best practice that you should be doing active learning in the classroom. Relative to when I was in school, you see a lot more teachers making students do things.”

Khan admits, however, that “we still are primarily in the seat time system” - this American reference to “seat time”, also known as the Carnegie Unit, refers to the awarding of marks on a minimum amount of instructional time. “It’s not as competency-based as it could be,” he says.

So, why hasn’t Khan Academy ushered in a new era of education?

Well, the platform is designed to help students who are “trying to get through the [public school] system”, Khan says. “Either we support them in moments where they have a gap or we are used more systematically by their teacher, by their school, to improve the learning that goes on.”

And the public school system is far bigger than Khan Academy, he says.

The academy “needs to be pretty well integrated with the formal systems for it to have the maximum impact. That’s the journey that will keep us busy for decades to come”, its founder says.

Salman Khan’s second attempt at an education revolution came about through circumstance: this time he would be the passenger, not the driver.

When Covid-19 forced schools to close, Khan Academy traffic tripled, “from 30 million learning minutes a day to about 90 million learning minutes a day”, Khan says.

Similar edtech products saw similar leaps in usage and the dominant narrative was that the tech revolution had finally arrived: schools had been forced to change, the benefits would be clear and there would be no going back.

Khan was excited, and ready to seize the moment. But it didn’t turn out how he or others expected. Rather than edtech tools being utilised to their full extent, what emerged was just a technologically facilitated version of school: lectures via Zoom, students working on shared documents as if they were the physical books they had always used in class. Pedagogically, nothing had changed.

A missed opportunity during the pandemic

“As soon as the schools started to implement their own virtual learning, we saw our users go down,” Khan says. “It’s obvious in hindsight, because of screen fatigue. And people were just trying to go with a minimum viable experience - they weren’t thinking about optimising.

“Teachers weren’t saying, ‘OK, let’s see if you can do personalised learning.’ No. They were just like. ‘Let’s just give the kids some semblance of school.’”

There was an opportunity for a revolution, but the circumstances weren’t right, Khan says. Worse, the chance for progress actually reduced.

“I think the pandemic - even from an edtech point of view - was a huge step back,” he says.

He puts this down to “the suddenness with which it happened” - schools weren’t prepared to immediately shift their systems overnight.

“People were just trying to do the basics. No one was planning on doing full virtual learning, and they had to shift. All the school systems tried their best, but they took a pared-down version of what was going on in the classroom, then Zoom-ified it. It’s never going to be good.”

And if any lessons were learned from the shift that the pandemic enforced, they didn’t transfer back to the classroom once schools reopened, Khan argues.

This was partly because the technology didn’t easily transfer, he explains. Some “thoughtful teachers” used breakout rooms to lead some “pretty stimulating classrooms on Zoom, sometimes more stimulating than might have been happening with the traditional lecture. But even those teachers, when they came back into the classroom, well - it’s hard to press a button and have every student go into their breakouts”.

What’s more, the pressures on schools to return to “normal” were just too great to think about a system transformation, he says.

“People felt like it was a victory just to come back. And I think we’ve been in that state since.”

Third time lucky? For Salman Khan’s latest attempt to revolutionise education, he has turned collaborator. And his focus this time is artificial intelligence (AI).

Khan estimates that he was “one of the first 20 people in the world” to use the fourth iteration of ChatGPT, ahead of its public release. And he explains that Khan Academy has worked with OpenAI to use its ChatGPT software on an extension of its existing platform.

In March 2023 Khan Academy launched Khanmigo, an AI function that acts as a “personal tutor”. Rather than simply answering questions, Khanmigo “facilitates Socratic dialogue”, Khan says, asking students why they think certain things or drip-feeding them with information to lead them to a solution, rather than simply giving answers.

While Khan Academy videos remain free, Khanmigo costs $4 per month for students using it at home, to cover “significant computation costs”, Khan says.

Meanwhile, its teacher function - offering analytics and tracking - is free to all educators in the US, and Khan says it will be rolled out, also free of charge, to “all English-speaking teachers” - including those in the UK - by the end of the summer, too.

“I don’t view AI as a replacement for the teacher”

While Khan seems certain of the transformative power of AI to finally change how we teach, he’s likely to face cynicism. He has long been criticised by teachers who regard his techno-optimism with suspicion.

An obvious fear here is that AI might come to replace teachers. But Khan strongly denies this possibility.

“I don’t view AI as a replacement for the teacher,” he says. “I view it as an augmentation. It gives the teacher more leverage. I don’t think AI will be better than a great human teacher, but there are things that AI can do that will help the human teacher.”

Specifically, AI could assist with administrative work, lesson planning and marking, he says.

“I met an English teacher in the US who told me she has 180 papers to grade on some weekends. That’s too much work. No matter how invested you are, if you’re reading 180 papers, it’s very hard to grade the 179th paper with the same fidelity that you grade the second paper.

“I’d still always put the teacher in charge, but the AI could help the teacher be dramatically more productive, and maybe even add an element of consistency,” he adds.

The AI revolution in schools?

Khan says AI could help at a classroom level, too. “A teacher can’t replicate themselves 30 times so that they can be with every student all the time,” he argues.

Now, however, a class of 30 students could “break out” into five groups, each around an “Alexa-type device” formatted with AI that facilitates discussion. “It wouldn’t even have to be every kid on a computer, because that’s very unnatural,” he explains.

While Khan insists that AI is to be used alongside - not instead of - teachers, he often refers to Khanmigo as a “teaching assistant”. This, too, could be seen to be problematic: TAs don’t want to lose their jobs either.

His understanding is based on the American system, where TAs aren’t as common at a school level. But once their role in the English system is explained to him, he says “it’s just the same principle” as with teachers - “they’re able to do more”.

Again, AI offers an opportunity for lessening the workload of staff, which frees them up to focus on in-person interactions with the children. “The more support and attention a student can have, the better.”

“The teacher tools have immediate, tangible effects that teachers hopefully feel energised by,” Khan adds. “And if you talk about teacher shortages, teachers leaving the field, teachers being undercompensated - well, if some of a teacher’s job becomes a little bit easier, that’s a win.”

School leaders currently working on AI projects - who wish to remain anonymous due to the early stages of their work - told Tes that AI has potential in their settings, but there is still a lot to be done in terms of fine-tuning these tools so they are truly effective.

Currently AI programmes can be a good starting point for lesson planning and for producing tailored resources for students with differing abilities, leaders say.

There are also hopes for the potential of predictive attendance software, which will add value to the work that trusts and schools are already doing to improve student attendance.

But there are concerns that the technology is not yet accurate enough to be relied upon. Khan’s belief that AI will help teachers with marking, leaders say, is not yet realised, because of a lack of capability to specify marking criteria.

What’s more, there are concerns around the number of AI products available, which leaves leaders unsure of which to invest in, particularly given the fast rate at which once cutting-edge technologies are being superseded by others.

So is Khan’s third attempt to transform education likely to end in the same way as his others? School leaders tell Tes that they are eager to embrace new technology - when and if it really helps.

Khan is confident the technology can get there, but - perhaps due to previous false dawns - he’s also more cautious this time around. AI is not the transformative agent but a step along the way, he suggests. He’s wary of absolutes.

“I won’t pretend,” he says, “that AI is a silver bullet.”

Ellen Peirson-Hagger is senior writer at Tes

For the latest education news and analysis delivered directly to your inbox every weekday morning, sign up to the Tes Daily newsletter