Applied Frontier

Regions · South Asia · Last updated November 2026

How can AI create new jobs in South Asia, and what should governments do now?

Direct answer

AI can create new jobs in South Asia because far more of the region’s work can be improved by AI than replaced by it. The World Bank estimates that about 15% of South Asian jobs are strongly complementary to AI, while about 7% are highly exposed with little scope for AI to assist the worker. The exposed work is concentrated in outsourcing and IT services, so the task for governments is to move those workers from producing output to verifying, deploying and owning it.

How exposed is South Asia to AI, and what is the real risk?

About 22% of jobs in South Asia are exposed to AI, according to the World Bank, and those jobs account for 42% of all wage earnings because they pay better. Agriculture still provides around 40% of regional employment, which AI reaches last.

The pressure falls on business process outsourcing and IT services. The World Bank reports that hiring in those sectors is slowing, that demand for call-center agents fell 24% relative to the least exposed occupations after late 2022, and that foreign firms may use AI to bring outsourced work back home.

The risk is concentrated, not general: the entry-level service jobs that built the region’s export sector are the ones AI replaces first.

15%

of jobs strongly complementary to AI (World Bank, 2025)

7%

of jobs highly exposed with low complementarity (World Bank, 2025)

~30%

wage premium for jobs requiring AI skills over other white-collar jobs (World Bank, 2026)

Why is this a once-in-a-generation window for South Asia?

Three things line up for roughly the next three years.

  1. 01

    Capital is arriving. In 2025 Google announced a $15 billion AI hub in Visakhapatnam, India.

  2. 02

    Demand is already visible. The share of South Asian job postings asking for AI skills rose from 2.9% to 6.5% between January 2023 and March 2025, and jobs requiring AI skills pay nearly 30% more than other white-collar work.

  3. 03

    The workforce is there. South Asia has about 1.17 billion people of working age, roughly 22% of the global labor force, and a large services workforce that already serves global clients.

The World Bank expects the region’s working-age share to peak around 2034. The decade in which South Asia has the people is also the decade in which the jobs are being redefined.

Which job families should South Asia build first?

Four of the eight job families a developing country can build in three years. None requires a four-year degree. These priorities are working hypotheses, to be tested in targeted pilots.

2–4 months on top of domain experience

Domain verifier

Checks AI output in finance, health, law and software before it is acted on. The natural next step for outsourcing and IT workers whose production tasks are being automated.

2–3 months to train

Agentic workflow builder for SMEs

Assembles AI agents into working processes for small firms. About 23% of surveyed Indian firms use AI, against 43% in the United States, and small businesses report difficulty identifying suitable tools.

2–6 weeks added to an existing role

AI-augmented frontline worker

Health workers, agricultural extension agents, teachers and clerks working with an AI copilot. Agriculture alone accounts for around 40% of regional employment.

3–6 months to train

AI infrastructure operator

Data-center operations, power, cooling and network, following the compute investment now arriving in India.

Where will these jobs cluster?

Around Super Intelligence enablement hubs: local institutions that help businesses and public agencies adopt advanced AI, give technology providers a local partner, and train the people who run the systems. In South Asia we expect the first hubs in India and Bangladesh, serving both national and regional markets.

How an enablement hub works →

What should a ministry in South Asia do this year?

  1. 01

    Protect the export base: retrain outsourcing and IT-services workers for verification, deployment and outcome ownership before their current tasks are automated.

  2. 02

    Stop funding new programs in routine coding, basic data entry and scripted customer support, and redirect the money to the job families above.

  3. 03

    Give small firms a route to adoption: shared cloud access, trusted advice and trained workflow builders.

  4. 04

    Fix the foundations AI depends on: literacy, numeracy, reliable electricity and internet. Adult literacy is around 75%, against 85% across emerging and developing economies.

  5. 05

    Negotiate skills commitments alongside every compute investment, not after it.

Where do multilateral agencies fit in South Asia?

The World Bank, the Asian Development Bank, UNDP and other multilateral agencies already fund and run education and workforce programs at scale. The change is in what they fund: outcomes-based funding for short, verified-skill programs tied to the job families above, with cost per placed worker as the unit of account. Applied Frontier is building the job-family map, the two lists and the readiness metrics that make such programs fundable, and runs the targeted pilots that prove a job family before a multilateral agency funds it at scale.

Unit of account

Placed worker, retained at 6 and 12 months

Program length

6 weeks to 6 months

Time to first data

One quarter

Work with us

Country reports

Which countries in South Asia have a country report?

India and Bangladesh. Each report is updated every year with the same indicators, so that jobs needed, AI exposure and cost per job can be compared across countries and over time. All country reports are listed under Regions.

2026 edition

India

7.85Mnon-farm jobs needed a year

Benchmark training cost of an AI-economy role: $240–$480

India country report →

2026 edition

Bangladesh

2Mlabor-market entrants a year

Benchmark training cost of an AI-economy role: $340–$680

Bangladesh country report →

What does each country report cover?

  • How many jobs the country needs each year, from population and labor-force data.
  • How exposed the country is to AI, and where the pressure falls.
  • The biggest untapped AI opportunities, by job family.
  • What it costs to create a professional today, and the benchmark cost of training for an AI-economy role.
  • What the government can do in the next 12 months.

Questions about AI and jobs in South Asia

Will AI take jobs in South Asia?
Some, in a narrow band. The World Bank estimates that about 7% of South Asian jobs are highly exposed to AI with little scope for AI to assist the worker, mainly in outsourcing and IT services. About 15% of jobs stand to become more productive.
Which South Asian jobs are most at risk from AI?
Call-center, customer support, routine software, accounting and payroll work. The World Bank reports that demand for call-center agents fell 24% relative to the least exposed occupations after late 2022, and warns that foreign firms may bring outsourced work back home.
What should South Asian governments train people for?
Roles that sit between a global model and a local outcome: domain verifiers, workflow builders for small firms, AI-augmented frontline workers and AI infrastructure operators. Each can be trained in weeks to months and credentialed by demonstrated work.

Sources

  1. World Bank, South Asia Development Update: Jobs, AI, and Trade (October 2025), summarized in “Is AI adoption impacting job markets in South Asia?”
  2. World Bank, South Asia Economic Update: Adopting AI for Growth (October 2026), as reported by The Kathmandu Post
  3. World Bank, South Asia Economic Update: Adopting AI for Growth (October 2026), as reported by Devdiscourse
  4. Google, “Our First AI Hub in India, Powered by a $15 Billion Investment” (2025)