How AI could transform Nepal's classrooms
From teacher shortages to outdated curricula, it presents scalable solutions
In a government school in rural Nepal, a student asks a chatbot what the Treaty of Sugauli was. The bot replies in crisp Nepali, provides historical context, suggests related topics for further exploration and suggests testing the student with follow-up questions. For a fleeting moment it seems Nepali education might leapfrog into the 21st century, not by importing foreign teachers but by renting synthetic ones.
The notion artificial intelligence (AI) can revolutionise classrooms is not new. But the proposition has become newly tantalising for countries like Nepal, where the education system is burdened by chronic teacher shortages, rote learning and threadbare infrastructure. If AI can provide personalised tutoring and help students learn at their own pace, why not plug the learning gap with silicon rather than wait for a teacher who may never arrive?
Officials at the Ministry of Education have shown growing interest in integrating artificial intelligence into classrooms, exemplified by the recent Mini-EdTech AI Conference they co-organised to explore AI’s role in education. Separately private initiatives are already in motion. During the COVID-19 pandemic, Fusemachines, a New York-based AI company with operations in Nepal, launched an Education Digitization Grant to support schools transitioning to online learning. Its Fuse Classroom platform, used by more than 25,000 students and teachers in Nepal and abroad, provides features like AI-powered chatbots, activity recommendations, automated grading and progress tracking.
It is not quite a digital utopia, but in a country where children still walk hours to attend understaffed schools, it has appeal. Yet enthusiasm ought to be tempered. As history shows techno-utopian dreams in education have a knack for overpromising and underdelivering. In the early 2010s “One Laptop per Child” initiatives were meant to transform classrooms in poor countries. Many of the laptops became paperweights. In Nepal a project in 2012 to deploy tablets in rural schools fizzled out due to connectivity issues, lack of training and—ironically—power cuts. The lesson is that hardware without hard thinking achieves little.
AI’s potential is real but so are its limits. Take personalised learning, one of the technology’s most hyped offerings. By tracking a student’s performance, AI can identify weak areas and tweak exercises accordingly. This works well for structured subjects like maths and science, where right and wrong answers are clear-cut. But in the humanities, nuance matters. A machine might guide a discussion on Shakespeare’s Macbeth, but it cannot replicate the spark of a good classroom debate—or the sideways glance between two students who realise they disagree.
Worse, generative AI tools sometimes hallucinate. They fabricate facts, cite non-existent sources or mangle logic. In a classroom setting such errors are not only inconvenient, they are corrosive too. If a student memorises nonsense fed by a chatbot, the damage may be hard to undo. Proponents argue these glitches are teething problems. But there is no guarantee better training data or more computational heft will fully solve them.
Even if the technology were flawless, motivation remains a human affair. Studies from America and Europe show self-paced learning platforms work only when students are already diligent. In Nepal, where first-generation learners usually lack support at home, and where school is as much a social institution as a pedagogical one, isolating students with screens may backfire. “Talking to a chatbot doesn’t help you stay in school,” says one education researcher in Kathmandu. “It’s the football match after class or the teacher who knows your father.”
Still, even sceptics concede AI could play a supporting role. Used judiciously it might lighten the bureaucratic burden on overworked teachers, many of whom juggle multiple grade levels with minimal preparation time. Already, early adopters in Nepal’s private schools use AI tools to create differentiated worksheets, translate lesson plans, automate exam grading. Such tools do not threaten teachers’ jobs but rather rescue their sanity.
Another promising use case is language. Government schools teach in Nepali but many students speak Maithili, Bhojpuri, Tamang or other mother tongues at home. AI-powered translation and voice tools could help bridge the gap, giving children access to content in their native language without requiring a national overhaul of curricula. It is a small step toward educational inclusion that does not require rewriting textbooks.
The bigger opportunity may be feedback. Tools like TeachFX in America analyse classroom recordings to tell teachers how much time they spend talking versus listening. Similar tools, adapted to Nepali classrooms, could nudge teachers toward more interactive styles. In schools where inspectors rarely visit, AI might even serve as a crude proxy for accountability—provided privacy concerns are addressed and local contexts respected.
Still, any effort to modernise education must reckon with the peculiar resilience of the system itself. Schools are often caricatured as dinosaurs: slow, conservative and maladapted to a fast-changing world. But as Justin Reich of MIT notes, they might better resemble great white sharks: largely unchanged for decades, not because of inertia but because their design still works. Even though age-graded classes, bells, blackboards and group learning may be old-fashioned they endure because they serve pedagogical and logistical functions that no app can replace.
Nepali policymakers would do well to remember that. The goal should be to make schools better at doing what they already try to do. That means making sure AI complements human teachers not supplants them; that it boosts social learning not isolates students; and that it lifts basic literacy and numeracy not only entertains bored children with talking robots.
It also means focusing on old problems as well as new ones. The country still lacks reliable internet in many villages. Teacher training is still woefully inadequate. Education budgets are tight, and procurement processes opaque. An AI-powered chatbot is no match for a corrupt official or a missing roof.
There is, however, one way in which AI might push schools to improve—by aspiration. Watching a bot explain geometry better than their textbook might inspire students to expect more. Watching teachers adapt and collaborate with new tools might restore some faith in the system. In that sense, AI’s biggest contribution to education may not be what it does but rather what it demands: that humans raise their game, too.
That is not a bad outcome. After all, unlike Thomas Edison’s motion pictures or 2010s tablet giveaways, AI has something those earlier tools lacked: users who talk back. When a student asks a chatbot whether education can really change their life, the most important answer may not come from the machine. It will come from what happens next. ■