ILLUSTRATION: SHUTTERSTOCK
IN SEPTEMBER 2025 Nepal’s streets filled with rage. Youth-led protests against corruption and a social-media ban met lethal police force. Human Rights Watch counted 19 dead. Prime Minister K.P. Sharma Oli resigned the next day. The grievances were many. Unemployment alone did not explain them. But a year later, artificial intelligence threatens to make the economic anger permanent.
Nepal was already failing its young. The World Bank found average real growth of 4.2% a year between 1996 and 2023, behind regional peers. Growth slowed to 3.9% in fiscal 2026 and is projected at 3.7% in fiscal 2027. Unemployment sits at 10.5%. Nearly one in five young men aged 15 to 24 is neither in education, employment nor training. Around 90% of workers are informal. Remittances reached 35.8% of GDP in fiscal 2026. The country does not create enough decent work. It exports its people and lives off the money they send home.
AI does not need to cause mass layoffs to inflict damage. It attacks the entry-level job. Customer support, bookkeeping, basic coding, translation, routine administration and junior analysis are exactly the tasks generative AI performs cheaply. The International Labour Organization estimates that one in four workers worldwide holds an occupation with some exposure to generative AI. Clerical work is particularly vulnerable.
The process is brutal in its simplicity. A Kathmandu accounting firm serving foreign clients once hired graduates to reconcile invoices and prepare routine reports. AI now does much of that. The firm needs fewer trainees. Experienced staff supervise the machine and become more productive. Profits rise. The graduate loses the chance to learn. This is career-ladder compression. Hire fewer beginners today, and the economy will lack experienced professionals tomorrow. More immediately graduates cannot get experience because the jobs that provided it have vanished.
The evidence from other countries is already ugly. A Financial Times report on October 8th described an estimated 5% global decline in call-centre employment since late 2023, using data from Revelio Labs and economist Caelan Wilkie-Rogers. The Philippines lost 10%. Kenya lost 6%. The adjustment came through reduced recruitment, not dramatic redundancies. Nepal should not wait for its own body count.
The ILO and World Bank working paper Disruption without dividend?, published in March, explains the trap. Developing countries have lower aggregate exposure to AI than rich ones. But they suffer disruption before they capture productivity gains. Digitally connected office workers are exposed quickly. Rural businesses that could use AI to improve farming or tourism often lack electricity, connectivity, finance and skills. Nepal could automate its better-paid jobs before it modernises the rest of the economy.
The outsourcing model is a death trap. For years developing countries sold cheap labour to foreign clients. If AI lets a foreign company do the same work with a smaller team, the Nepali supplier loses the contract. A handful of specialists may remain to manage complex cases. They will not absorb the displaced workforce. Competing on low wages against a rival armed with AI, better processes and proprietary expertise is a losing game.
Productivity can rise while employment stagnates. A firm doubles output with 20% more staff. The owners and skilled workers capture the gains. The young are left outside. The World Economic Forum expects 170m jobs created and 92m displaced globally by 2030, a net increase of 78m. That is a global projection and not a promise for Nepal. New jobs do not appear automatically where old ones die.
The political consequences are combustible. The September 2025 protests showed what happens when youth anger meets corruption, heavy-handed policing and a trigger. AI could add economic despair to that mix. Educated young people are dangerous when they believe the system has cheated them. Families sacrifice for degrees. Governments promise development. Employers demand experience. If none delivers, the social contract rots. Social media lets the frustrated compare notes and organise. The result can be protests, instability or worse.
Politicians will be judged by jobs and not by training courses. Telling graduates to learn AI prompts is a confession of impotence. Nepal needs paid apprenticeships, demand-led skills, support for small firms to adopt AI, better infrastructure, access to finance, export markets and portable social protection. It needs an employment observatory to track vacancies, starting salaries and migration. It needs entrepreneurs who can build firms, not freelancers competing for shrinking scraps.
None of this guarantees success. AI can help a trekking company reach foreign customers, a farmer monitor crops, a small software team build specialised products. But productivity without broad opportunity is a recipe for revolt. If Nepal fails, it will become a waiting room for the young, a country that exports its talent and imports its regret. The ladder is being sawed off. The government can build a new one, or watch the fall. ■
