Khan Academy built an AI tutor called Khanmigo in 2023. By 2025, it was being used by millions of students in India and sub-Saharan Africa who could not afford private tutoring. The students who had access to a patient, always-available, infinitely adaptive tutor for the first time were not replacing their teachers โ they were closing the gap between what their teachers could give them in a 40-student classroom and what they actually needed. This is what AI in education actually looks like, and it is more hopeful than the discourse suggests.
What AI Is Doing in Classrooms Today
Personalised tutoring at scale is the most transformative development. AI tutors can explain the same concept 10 different ways, identify exactly where a student's misunderstanding is, and pace themselves to the learner. For students in Nairobi or Hyderabad who share a teacher with 45 peers, this is access to a quality of learning support that was previously available only to families who could afford private tutors.
Automated grading and feedback is in active deployment across Indian schools and increasingly in East African institutions. Subjective essay grading remains hard for AI, but MCQ, short answer, and coding exercises can be graded instantly with detailed explanatory feedback. Teachers who used to spend weekends marking assignments get that time back.
Curriculum planning and material generation โ teachers using AI to generate lesson plans, create differentiated materials for different learning levels, and adapt content to local context. An English teacher in Nairobi can ask an AI to rewrite a comprehension passage using examples from East African business news rather than British literature, and get a usable result in seconds.
What AI Cannot Do in Education
The honest answer is that almost everything that matters most about a great teacher is beyond AI's current reach. A teacher who notices that a student who was engaged last week is withdrawn today โ and thinks to ask what is happening at home. A teacher who makes a child feel seen and capable for the first time. A teacher who models how a curious adult engages with the world.
These are not minor footnotes to education โ they are much of what education actually is. The schools and systems that treat AI as a replacement for teachers rather than a tool to extend them are going to learn this the hard way. The ones treating AI as infrastructure โ handling the mechanical load so teachers can do the human work โ will produce better outcomes.
What Educators Should Do About AI
Use it to eliminate low-value work. Generating first drafts of materials, grading routine assessments, creating differentiated practice sets. Reclaim the time AI saves and reinvest it in the high-value work: student relationships, mentoring, discussion, the parts that require a human in the room.
Learn to teach AI literacy. Students in India and Africa are going to graduate into a workforce that requires AI fluency. Educators who understand AI well enough to teach it thoughtfully โ including its limitations and failure modes โ are providing something increasingly essential. This does not require becoming a programmer. It requires staying curious and willing to learn alongside your students.
Advocate for equitable access. The risk with AI in education is not that it replaces teachers โ it is that it widens the gap between students with access to good AI tools and those without. The educators and administrators shaping how AI gets deployed in their schools have an opportunity and a responsibility to ensure that access is built into the design.
Learn AI the way it should be taught
Ramayai is built on the same philosophy โ real projects, real feedback, at a pace that works. Our AI Agents track is the place to start if you want to understand what you will be teaching.
Explore the AI Agents Track โ