Most of the conversation around AI in education tends to focus on the same things: students using it to cheat, teachers trying to detect it and universities rewriting policies faster than students can find ways around them. It’s a real conversation and a fair one, but it’s also only half the picture.
The other half is quieter – it’s founders and teachers who have actually seen where education struggles and built something to address a specific problem. They are using AI for feedback, admin, personalisation and practice, the less flashy parts of education that can still have a big effect on how someone learns. So rather than asking whether AI belongs in a classroom, it’s maybe better to look at what it’s actually being asked to solve.
AI Doesn’t Have To Do The Learning For Students
There is a very clear difference between a student getting an answer and a student being taught. Felipe Hildebrand, co-founder and CEO of Neomind Learning, says it quite simply: “A child who receives an answer has finished a task. A child who has been taught has changed. Those are different outcomes, and only one of them matters.”
Hildebrand also points to this year’s PISA results, noting that “math and reading scores have declined by 22 and 28 points respectively over the past decade” while “students using AI daily for schoolwork scored around 25 points below those who never do.” He makes sure to note that this is correlation rather than causation. His argument is that AI has often been pointed at the wrong problem.
Rather than just producing an answer, Hildebrand believes AI could be used to understand the learner first – where a gap appeared, what has been misunderstood and how that particular student learns best.
Felix Ohswald, CEO and co-founder of GoStudent, is approaching a similar problem from the tutor’s side. Stu, GoStudent’s AI co-tutor, supports tutors with lesson preparation, exercise generation and resource recommendations. That can take some of the work around a lesson off the tutor’s plate, which would leave more time for the actual student.
Outside lessons, Ohswald says AI can also help explain difficult concepts, summarise information or approach a topic differently. The point is to give students another way to work through something they are struggling with instead of completing the work on their behalf.
What Could Personalised Learning Look Like?
Most teachers would probably like to give every student more individual attention. The problem is that there are only so many hours in a school day, which is something Mayank Kejriwal, AI professor at USC and co-founder and CEO of GRAIL, is trying to address.
GRAIL’s agentic teaching assistant is being used in a graduate class at RPI, where it supports around 16 students between instructor touchpoints. Kejriwal says the instructors have found they can give more detailed, personalised feedback within a week, compared with at least two weeks in earlier semesters. As he explains, “The best systems should not simply hand students answers. They should help students practice, get feedback two to three times faster on homework and assignments, and personalise to existing abilities.”
Rather than treating AI as an automation tool, Kejriwal sees it as a way to give instructors more time for the parts of teaching that need their judgement. As he puts it “AI can give instructors time back for mentoring and personalised feedback”, allowing them to focus more on the individual needs of their students.
Penelope Barton, co-founder and CEO of Crimson Global Academy, approaches the same problem from a different angle: visibility. “If 20 students ask questions about equations, for example, that information can help a teacher identify a shared gap and shape the next lesson accordingly”, she explains.
Barton is clear that this doesn’t replace the teacher. Instead, AI can give them more information about what happened between the lesson and the submitted work. It could show where a student became stuck, what sparked their curiosity or which concepts maybe need more attention.
AI Can Give Students More Ways To Practice
Dr Panos Siozos, CEO and co-founder of LearnWorlds, says that “Agentic AI can personalise learning and facilitate real-time, real-world scenarios and simulations where students can practice and test decisions.”
For example, Siozos points to a law student negotiating against an AI agent that shift its position, or a language learner practicing with a reflexive AI agent. It gives students more room to practice, make mistakes and try again while they’re still learning. Students can make mistakes while they’re still in a learning environment, rather than discovering the same gaps when the stakes are much higher.
Colin Coggins, CEO of Chegg Skills, looks at this from the skills side. He argues that education needs to give students more opportunities to apply what they know to unfamiliar problems, receive feedback and improve.
“56% of employers say entry-level workers aren’t adequately prepared for work,” he says. For Coggins, thoughtfully designed AI could create more opportunities for students to practice that cycle rather than fewer.
The focus is less about getting students through a task and more about giving them somewhere to work through difficult skills before they have to use them for real.
Could AI Be The Missing Piece In Modern Education?
Put together, these aren’t all versions of the same idea – they show several different ways AI can support education: helping students understand rather than handing them an answer, giving teachers more time for feedback, taking some work off their plates and creating more opportunities to practice difficult skills.
None of that means AI will suddenly fix education; it does however, suggest that the more valuable conversation is about what AI is actually being built to solve. When it gives teachers useful information, takes care of time consuming work or gives students somewhere to practice, it could basically just become another tool (if we finally accept it) for supporting learning without taking the human part out of it.
Our Experts:
- Marlee Strawn: Co-founder of Tampa-based Scholar Education
- Mayank Kejriwal: USC AI professor,Co-founder and CEO of GRAIL
- Felipe Hildebrand: Co-Founder and CEO of Neomind Learning
- Felix Ohswald: CEO and Co-Founder of GoStudent
- Dr Panos Siozos: CEO and Co-founder, LearnWorlds
- Colin Coggins: CEO at Chegg Skills
- Penelope Barton: Co-founder and CEO of Crimson Global Academy
Marlee Strawn, Co-founder of Tampa-based Scholar Education
“I saw special education evaluation waits stretch from months to a full school year because schools didn’t have enough qualified staff and because of the paperwork burden.”
“Scholar isn’t replacing teachers or evaluators. We’re building AI infrastructure that gives back the hours lost to documentation so humans can focus on relationships, judgement and parent conversations.”
“The most useful AI in education is often the AI working quietly behind the scenes, not answering homework questions, but giving adults enough time and information to do their jobs well.”
“We’ve supported roughly 18,235 students across 29 Florida schools, helped deliver close to 39,481 IEP and 504 accommodations and estimated 4,137 hours of teacher time returned.”
Mayank Kejriwal, USC AI professor, Co-founder and CEO of GRAIL
“The public conversation around AI in education focuses heavily on cheating, but the more constructive question is how AI can deepen learning rather than shortcut it.”
“The best systems should not simply hand students answers. They should help students practice, get feedback two to three times faster on homework and assignments, and personalise to existing abilities.”
“GRAIL recently launched an agentic teaching assistant, piloting at RPI in fall 2026 in a graduate class, supporting around 16 students between instructor touchpoints.”
“Instructors are finding they can give more detailed, personalised feedback within a week compared with at least two weeks in earlier semesters.”
“AI can give instructors time back for mentoring and personalised feedback.”
Felipe Hildebrand, Co-Founder and CEO of Neomind Learning
“A child who receives an answer has finished a task. A child who has been taught has changed. Those are different outcomes, and only one of them matters.”
“Math and reading scores have declined by 22 and 28 points respectively over the past decade, roughly a year to a year and a half of schooling lost. PISA also found students using AI daily for schoolwork scored around 25 points below those who never do. That is correlation, not causation.”
“AI in education is often optimised for speed rather than learning. It can instead understand learner gaps, misunderstandings and how an individual learns.”
Felix Ohswald, CEO and Co-Founder of GoStudent
“AI can reduce teacher and tutor time spent on preparation and admin, including lesson planning, exercises and resource adaptation.”
“Stu, our AI co-tutor integrated into the virtual classroom, supports tutors with lesson preparation, exercise generation and resource recommendations.”
“Students can also get support outside school or tutoring hours, with explanations, summaries and different approaches to difficult concepts.”
“AI should help students understand, not simply do the work for them. It can complement existing teaching when students are struggling with difficult concepts or need a different way of understanding something.”
Dr Panos Siozos, CEO and Co-founder, LearnWorlds
“AI should support educator expertise rather than bypass it. The educator becomes the director of the learning experience, supplying intention, interpretation, context and standards while AI helps build the learning environment.”
“Agentic AI can personalise learning and facilitate real-time, real-world scenarios and simulations where students can practice and test decisions.”
“A law student could negotiate against an agent that shifts its position. A language learner could practice with a reflexive agent.”
“Teachers struggle to provide enough feedback. With one teacher and 30 students, useful comments may arrive two weeks late. AI can respond while the work is still fresh and let teachers focus on the judgement calls.”
Colin Coggins, CEO at Chegg Skills
“AI is exposing an education crisis rather than creating it. Education has measured what students know more than what they can do.”
“56% of employers say entry-level workers aren’t adequately prepared for work.”
“AI can remove friction around the teacher and learner, provide immediate feedback, personalised practice and support at scale, while freeing educators for judgement, connection, motivation and mentorship.”
“Students need more opportunities to practice, apply what they know to unfamiliar problems, get feedback, improve and prove their capability.”
“The shift AI can enable is from ‘Did you learn it?’ to ‘Can you use it?'”
Penelope Barton, Co-founder and CEO of Crimson Global Academy
“AI changes education because information is faster and cheaper, but an answer isn’t learning.”
“If 20 students ask questions about equations, for example, that information can help a teacher identify a shared gap and shape the next lesson accordingly.”
“AI can show where a student got stuck, what sparked curiosity and which concepts need attention.”
“AI should give skilled teachers better visibility into the learning journey.”
“Answers are now cheap and fast. Capability is not.”
“A brilliant teacher using AI thoughtfully will outperform either one alone.”
