If you've heard about students and generative AI lately, you may have heard a few discussions about the adoption of tools like ChatGPT. Are they helping? (Yay! Great for research! Fast!) Or hurting? (Boo! Misinformation! Cheating!) But some startups are embracing the introduction of generative AI into school environments in a positive, and understandable, way. And they're building products to address what they believe to be a market opportunity.
Now, one of them has raised funding to realise that ambition.
MagicSchool AI, which is developing generative AI tools for educational environments, closed a $15 million Series A round led by Bain Capital Ventures. Denver-based MagicSchool started out making tools for educators, and now about 4,000 teachers and schools use the company's products to plan lessons, create tests and develop other learning materials, founder and CEO Adeel Khan said in an interview.
More recently, it has also begun building tools for students through schools. MagicSchool plans to use the funding to continue building further on both fronts, as well as to acquire more customers, hire top talent, and more.
This latest round also includes backing from some very prominent investors, including Adobe Ventures (whose parent company Adobe is very committed to AI on its platform) and Common Sense Media (a specialist in age-based tech reviews that is moving into generative AI with its AI guidelines partnership with OpenAI and chatbot evaluations). Private investors in this round include Replit founder Amjad Masad, Clever co-founders Tyler Bosmeny and Rafael Garcia, and OutSchool co-founder Amir Nathoo (several of whom were also seed investors in the company, which previously raised around $2.4 million).
Khan didn't disclose Magic School's valuation in this round, but investors believe backing investments in applications like this one is a natural next step for the AI startup after pouring hundreds of millions of dollars into infrastructure companies like OpenAI, Anthropic and Mistral.
“AI is coming to education, and there's a huge opportunity to build assistants for both teachers and students,” Christina Melas Kyriazi, a partner at Bain Capital Ventures, said in an interview. “There's an opportunity for teachers to help them with lesson planning and other tasks that they do remotely.”
From teacher to AI evangelist
MagicSchool, despite its name, didn't appear out of thin air.
Khan began his career as an educator, working with Teach for America after college (although his interest in public service and the role of education may have preceded that: he was student body president at Virginia Tech at the time of the shooting, giving him an unfortunate front row seat to the devastation of gun violence).
As a teacher, he showed early signs of unearthing both an entrepreneurial spirit and an interest in leadership when he moved to Denver with the idea of starting his own school.
He served in various administrative roles in local schools before eventually founding his own charter high school, Conservatory Green High School (DSST), which saw 100% of its first graduating class get accepted into a four-year college.
During a break from his hectic career, Khan came up with the idea for a magic school.
“It was around November 2022 when ChatGPT started dominating the news headlines and generative AI became a hot topic in much of the country,” he recalls. “I started tinkering with it as I was thinking about what to do next, and I quickly realized how useful this new technology could be for educators.”
He workshopped early versions of using generative AI to build tools for teachers, visited schools where he had taught, and talked to former colleagues about the possibilities. But it didn't take off.
“The interface was clunky and difficult to use,” he said. When Khan showed teachers a demo, they elicited the expected “wow” reaction, but left them alone, they would try it once and then never use it again.
“They told me, 'I spent so much time telling the robot what to do that it did what I wanted it to do that it ended up wasting time instead of saving me time.'”
His solution was to come up with more specific customizations.
“Behind the scenes, we were running some very sophisticated prompts to make sure the output was what educators expected,” he said.
Some examples of what teachers are creating with MagicSchool include lesson plans, quizzes and tests, course materials, and reorganization of materials for different levels of difficulty. MagicSchool is constantly improving all of this, and Khan says they are making heavy use of APIs from OpenAI, as well as others like Anthropic. Behind the scenes, they're doing AB testing to determine what works best in which scenarios.
Still, convincing teachers who don't pay for the product, or schools that do, to join MagicSchool has never been easy.
“When we started the product, we couldn't meet with any schools or districts, including the one I worked at, because there was so much uncertainty about it all,” he said. All it took to end the conversation was “negative headlines about the use of AI in schools, headlines about AI taking over the world and robots.”
That has slowly begun to change as society and industry adopt AI more widely and more advanced models are deployed. Saving time is the most obvious reason to use AI, he said, but he also finds it useful for brainstorming ideas and supplementing what he can learn on his own.
“I think educators didn't really know or have high expectations of what AI could do for them or their audiences,” he said.
Plus, he makes a second argument for why it makes sense to bring more AI into the classroom: AI will be part of how we do everything, and it's schools' job to prepare students for it.
AI is smart, but not as smart as humans
That said, there are limitations to using AI in every scenario, including the classroom.
“AI has a very different type of intelligence than human intelligence. Humans have evolved an emergent intelligence that is the product of millions of years of pruning by natural selection. It's very holistic and cognitively very flexible,” said Muturu Cukurova, professor of education and AI at University College London, which has a long-running lab that studies different combinations of AI and learning. (One very practical conclusion from the recent paper is that we need a hybrid approach that embraces both AI and humans.)
“AI has designed intelligence, not emergent intelligence, which means it is designed for a very specific goal, or set of goals. AI excels at this specific goal and shows notable signs of intelligence, but it is a different type of intelligence.”
This could be especially relevant for teachers who may not have enough experience with students and how they will learn in an AI world, or to judge whether AI versions of learning materials like quizzes might not be good enough.
Cukulova said that while automating certain tasks can be a valuable use case, “it becomes problematic when teachers don't have enough experience before they learn how to do these things themselves.”
Khan said Magic School aims to be considerate of this, especially when it comes to students. He said schools control what features they offer to students on the platform, and it's clear if they've used Magic School for assignments.
This all sounds great in theory, but only stress testing may ultimately reveal the cracks.
For example, will cash-strapped school districts try to rely more on input from AI systems rather than in-person time with teachers, or how can schools identify when students are using AI tools outside of the classroom without teacher approval?
This will require a different kind of AI education, Cukulova said: “This is a key piece of the puzzle: How do we educate and train people to use AI effectively and ethically?”