IMAGE: SHUTTERSTOCK
NEPAL HAS spent decades exporting its people. Artificial intelligence offers a chance to export more of their work instead. A programmer in Kathmandu can serve an Australian company without boarding a plane. A small firm can analyse medical records for an American client, process mortgage applications or build software for customers thousands of kilometres away. AI could make these services cheaper to produce and easier to flog.
The opportunity is already there. Nepal’s information-technology service exports may have crossed $1bn in 2025, according to the Nepal Association for Software and IT Services Companies. The estimate is built on industry guesswork and growth assumptions, not a comprehensive official tally. The last major study put ICT exports at about $515m in 2022. The distinction matters: Nepal has a promising export industry, but it still lacks the damn data to measure its full size.
AI could help that industry grow. It could also eat its lunch.
For years Nepal’s pitch to foreign clients has been brutally basic: educated workers at lower wages than those in rich countries. Companies provide software development, data processing, back-office services at competitive prices, etc. AI attacks the weakest part of that pitch. If a machine can complete routine coding or document processing, the client has less reason to pay a human team by the hour.
The mistake would be treating AI as a tool for making the same low-value work slightly cheaper. That invites a race against machines whose costs keep crashing. The stronger strategy is to sell the work that becomes more valuable when AI handles the routine parts: specialised knowledge, reliable judgement, domain expertise and responsibility for the final result.
A Nepali company processing mortgage documents for an Australian lender, for example, could use AI to extract information and flag inconsistencies. Its value would then depend on how accurately it verifies records, handles exceptions and meets the client’s compliance requirements. The product becomes a reliable outcome rather than a batch of labour hours. That is a difficult business to replace.
The same reasoning applies to healthcare data, accounting, engineering and legal support. Nepal already has firms serving overseas customers in some of these fields. AI gives them a chance to handle more work with the same staff, move into more demanding assignments and raise revenue per employee. It also makes mistakes more expensive. A fabricated entry in a medical record or a missed clause in a financial document can cost far more than the labour saved.
The second opportunity lies closer to home. Nepali businesses often struggle with slim margins, fragmented records and inefficient administration. A hotel could use AI to answer routine guest queries and adjust prices. A distributor could forecast demand and reduce inventory rotting unsold. A farm supplier could translate technical advice into local languages. A small exporter could prepare product descriptions and customer responses for overseas markets.
These uses are less headline-grabbing than building a Nepali version of a frontier AI model. They are also more likely to produce handsome gains. Most local firms lack the data, capital and specialist staff needed to train their own models. Many can still gain from existing tools, provided they redesign workflows rather than buy subscriptions and wait for miracles.
The government’s plans point in the right direction. Its policies for fiscal year 2026-27 prioritise AI exports, digital infrastructure, remote work and technology-driven employment. In March the Asian Development Bank approved a $40m concessional loan for Nepal’s digital transformation, alongside $50m from the World Bank for public-service digitalisation and supporting infrastructure. The ADB announcement focuses on government systems, digital services and so on. Such infrastructure can help businesses too, if implementation delivers.
Yet Nepal’s biggest constraint may be its rotten institutions rather than its technology. Companies need stable tax rules, reliable electricity and straightforward ways to receive foreign payments. They need access to global clients and employees who can combine technical skills with industry knowledge. The diaspora can help open doors, but relationships alone will rarely build a scalable export business.
Education needs to change as well. Training thousands of young people to write basic code could become a piss-poor investment if AI automates much of that work. Students need to learn how to test outputs, manage data, identify errors and understand the industries their software serves. Employers need incentives to train existing staff instead of treating AI as a shortcut to immediate headcount cuts.
There is a distributional risk. Larger companies can buy better tools and hire specialists to deploy them. Small firms may fall behind. Entry-level workers could lose the routine tasks through which they once gained experience. If AI raises output while weakening the first rung of the career ladder, Nepal could export more services while creating fewer opportunities for young graduates.
The country therefore needs a better measure of success than the number of people who learn to use a chatbot. It should track export earnings, revenue per employee, productivity gains and the quality of jobs created. It should also make it easier for firms to test AI on real business problems, with clear rules for, say, the handling of sensitive information.
Nepal’s advantage will never be the cheapest worker in the world. There will always be another country willing to work for less, and another machine able to do more. Its opportunity is to help businesses abroad solve difficult problems with a combination of human expertise and artificial intelligence.
The export product of the future may carry fewer hours of Nepali labour inside it. It should carry more value. ■
