AI, Employment and the Distribution of Development Gains in South Asia

 

According to the World Bank’s Bangladesh Poverty and Equity Report, an estimated 2.1 million people fell into poverty in the 2025-26 fiscal year due to a weak labour market and high inflation. At the same time, the country’s real GDP growth stood at 3.4 per cent. The organisation’s model-based estimates suggest that the national poverty rate may rise from 18.7 per cent in 2022 to 22.5 per cent in 2026. These two figures are calculated using different periods and methodologies. Therefore, it would be inaccurate to conclude that growth itself has increased poverty. However, a question remains: why is people’s living standard not improving at the same pace as the economy expands?

Without growth, it is difficult to reduce poverty, expand education and healthcare or finance social protection. However, an increase in per capita income alone does not provide a complete picture of people’s living conditions. Amartya Sen (1999), in his capability-based concept of development, showed that income is a means of achieving the life one chooses, not the ultimate goal of development. Education, health, security and the ability to make one’s own decisions are also part of development. Therefore, even when income increases at the same rate, people’s lives do not change equally. Savings, the nature of employment, household care responsibilities, access to markets and the realities of accessing public services determine who can turn additional income into real opportunities.

According to the World Bank’s South Asia Economic Update, the region’s growth is projected to be 6.9 per cent in 2026 and 6.7 per cent in 2027. The regional estimate in this report includes Bangladesh, Bhutan, India, Maldives, Nepal and Sri Lanka. Country-level forecasts vary considerably. India’s growth is projected to be 7.1 per cent in the 2026-27 fiscal year, Bhutan’s 6.7 per cent, Nepal’s 3.7 per cent and Bangladesh’s 3.4 per cent. The calendar-year forecasts for Maldives and Sri Lanka in 2026 are 2.2 per cent and 4.4 per cent, respectively.

Differences in economic structures are a major reason for these disparities. Construction of hydropower projects is contributing to growth in Bhutan, while Nepal is benefiting from hydropower exports. India’s electronics and machinery exports have offset much of the weakness in other sectors. Bangladesh’s situation is different. According to the World Bank, industrial production grew by around 4.4 per cent in 2025, but based on data for the first five months of 2026, the rate fell to around 2 per cent. Nearly 85 per cent of the country’s merchandise exports depend on the readymade garment sector. As a result, a decline in demand in European markets puts pressure on exports and related employment. Since the conflict in the Middle East began, tourist arrivals have fallen by around 13 per cent in Maldives and 9 per cent in Sri Lanka, while they have increased in Nepal and Bhutan. These differences show that a country’s ability to withstand shocks depends on market diversification, infrastructure and alternative income opportunities.

If energy prices remain high for an extended period, production, transportation and fertiliser costs increase. Agricultural production may also be disrupted by the effects of El Niño, potentially driving up food prices. Affluent households can cope by cutting spending in other areas, but poorer families may have to reduce spending on food, healthcare or their children’s education. The same price increase, therefore, does not affect everyone equally. People’s savings, alternative sources of income and social protection determine how much they suffer from a crisis.

Around 40 per cent of South Asia’s workforce is engaged in agriculture. A large proportion of them work in small-scale and informal employment. Increasing agricultural production is necessary, but new employment opportunities can also be created in post-harvest storage, processing, warehousing, transportation, wholesale trade and food services. A World Bank analysis found that hourly wages in intermediate and downstream activities in agricultural value chains, such as processing, transportation and retail, are 19-29 per cent higher than those in agricultural work. Reliable electricity, roads, cold storage, market connections and simplified regulations are needed to benefit from these activities. However, creating jobs alone is not enough. Workers must also have pathways to training, experience and better-paid employment. Farmers and informal workers need skills training, recognised certificates and job-search assistance so that they can move into new occupations without taking major risks of losing their income.

AI can increase productivity, exports and the quality of public services. However, access to technology remains highly unequal. In a 2025-26 World Bank survey, 23.4 per cent of formal firms in India reported using AI in some capacity. In the United States, the rate was 42.7 per cent. No other South Asian country was included in the survey. Therefore, India’s rate cannot be taken as representative of Bangladesh or the entire region. Large firms can invest in technology and train employees more easily, while smaller firms face barriers such as limited finances, skills and reliable internet access.

Another point worth noting is that adopting AI does not guarantee a significant increase in a firm’s productivity. In the World Bank survey, nearly three-quarters of Indian firms expected little or no productivity gains over the next three years from rapid AI development. Only 8 per cent of firms expected gains of 20 per cent or more. Small and medium-sized firms anticipated lower gains than larger firms. Around 72 per cent of Indian firms said improved digital infrastructure would help expand AI use, while nearly 60 per cent highlighted the need for better higher education. This means that lowering the cost of technology alone does not create equal opportunities for everyone. Skilled workers, quality management, reliable electricity and internet, and the capacity to reorganise work are also necessary. If these capabilities are lacking, both access to technology and the benefits it generates may remain concentrated among a limited number of firms.

Neither excessive optimism about AI nor fears of job losses provide a complete picture. According to the World Bank, around 7 per cent of jobs in South Asia are highly exposed to AI, where human skills are less complementary to the technology. The corresponding rate in other developing economies is 15 per cent. Meanwhile, 15 per cent of workers in South Asia are employed in jobs with high AI exposure where human skills are complementary to the technology, creating opportunities for productivity gains.

Exposure to AI does not necessarily mean job loss. A 2025 study by the International Labour Organization also found that most occupations are more likely to undergo changes in the nature of work than to disappear entirely (Gmyrek et al., 2025). However, workers will not automatically benefit from these changes. Acemoglu and Restrepo’s (2019) analysis identifies two opposing effects of technological change. Automation can reduce labour demand by transferring some tasks previously performed by humans to machines. On the other hand, the creation of new technology-related tasks can generate new demand for workers. Ultimately, the impact on labour income and employment depends on the balance between these two processes. Therefore, alongside training workers, it is also necessary to examine whether firms are creating new tasks that make use of human skills.

Whether technological gains reach workers depends on training, the creation of new tasks and bargaining power. A firm may use AI to make its employees perform more work. Alternatively, it may reduce their workload and allow them to focus on more complex tasks. However, higher productivity does not guarantee increased wages or job security. Workplace rules, wage-setting mechanisms, working hours and employee representation are important in this regard. If the benefits of technology are reflected only in firms’ output or profits, the calculation of its social gains remains incomplete.

Another World Bank analysis found that, following the release of ChatGPT, buyer relationships between AI-exposed US industries and South Asian supplier firms increased relatively rapidly. A one-standard-deviation increase in AI exposure in US industries was associated with a 27.5 per cent higher rate of growth in buyer relationships with South Asian firms. This indicates potential market opportunities. However, the study measured the number of relationships, not the financial value of each relationship. Larger exporters were also relatively more represented in the sample. Therefore, these opportunities cannot be assumed to reach small and medium-sized firms or all exporters in Bangladesh equally. Local firms need to ensure skills, quality, infrastructure and timely delivery to capitalise on these opportunities.

A controlled trial of AI use in healthcare in Bangladesh found that AI-assisted detection of diabetic eye disease at an eye care centre in Rangpur increased the number of patients served per specialist-hour by 40 per cent. The number rose from 1.14 to 1.59 patients per hour (Abramoff et al., 2023). Meanwhile, a six-week trial involving students at government primary schools in Quetta, Pakistan, found that teacher-led targeted lessons helped compensate for learning losses. However, the group using AI with teacher support did not show statistically significant improvements in outcomes. This was an early-stage experimental study, and its results cannot yet be generalised to larger settings or different environments. Nevertheless, both examples show that the outcomes of technology depend on the nature of the work, the role of people and the practical infrastructure available.

When technology reduces the time required to complete a task, it is also important to examine who ultimately benefits from that saving. Are workers required to do more work for the same wages, or are they given opportunities to increase their income, rest and develop skills? Whether workers control the time saved or employers use it to set higher work targets depends on workplace rules and bargaining power. Therefore, alongside productivity, actual working hours, additional workload, access to leave and workers’ autonomy over their schedules should also be considered. If technology saves time but workers do not have control over how that time is used, its benefits in their lives may remain limited.

Unequal distribution of time is not confined to the workplace. Long waits for public services, commuting, power outages and repeated visits to offices for approvals also cost people time and income. For those with little savings, a day away from work means a direct loss of income. Therefore, the time required to access a service, the number of visits and the associated costs should also be considered measures of administrative efficiency.

The proportion of working-age people in South Asia is expected to peak around 2034. By 2049, the share of people aged 65 and above may reach 14 per cent. Yet only 33 per cent of the workforce contributes to pension schemes, while just 10 per cent of older people receive pension benefits. This means the region must simultaneously increase employment and productivity while building financial security for old age. If today’s young people do not have access to stable jobs and opportunities to save, their own old age may also remain insecure.

For Bangladesh, the first priorities are reliable electricity, affordable internet, reducing barriers to banking and investment, and strengthening the revenue base. In agriculture, storage and market connectivity need to be improved. Tourism should generate local employment, while small firms need support to adopt technology. AI training must be linked to practical work so that workers can increase their income and skills by using new technologies. Low-cost technologies in local languages can initially be introduced for farmers, teachers and healthcare workers. However, the outcomes of each intervention, data security and accountability for erroneous decisions must be independently assessed. Regular publication of information and mechanisms for addressing complaints are also necessary to determine whether these policies are working in practice.

Social protection also has significant gaps. According to the World Bank, nearly half of the poorest households are not covered by any social protection programme. Subsidies and assistance should be designed so that limited public funds reach those who genuinely need them. The organisation’s analysis suggests that consolidating multiple food subsidy programmes and improving the targeted distribution of existing cash assistance through family cards could potentially lift an additional 2.85 million people out of poverty. This is a potential outcome dependent on policy changes and effective implementation, not a guaranteed forecast. Therefore, identifying the right households and regularly updating beneficiary lists are also essential.

Author: Dr Md Abu Hasan, Associate Professor, Shaheed Buddhijibi Government College, Rajshahi.

Leave a Reply

Your email address will not be published. Required fields are marked *