AI reaches India through two doors: Indian firms adopting it at home, and foreign clients adopting it for work they now send to India. This explorer runs Anthropic’s economic scenarios model, adapted for India, so you can see what your expectations about AI imply for growth, wages, jobs and who gains.
An independent adaptation of Korinek, Jones, Sacher, Cotter and McCrory (2026), Economic Scenarios for Transformative AI, built with Claude. Not an official Anthropic publication. Full method and sources in the technical report.
The original model describes a closed economy. India is not one. About a quarter of Indian workers are in occupations AI can reach, and they earn around 40% of labour income. A further slice of India’s cognitive work, tech and business-services exports of USD 246 billion, is done for foreign clients.
When Indian firms adopt AI, India gets the productivity gain and the displacement. When foreign clients adopt AI for work they used to offshore, India gets only the displacement. And part of the income earned by AI capital flows to its mostly foreign owners.
Each scenario pairs one of the original US scenarios with an Indian counterpart: the same technology, slower adoption by Indian firms, and export demand driven by the US path. Values are for the start of 2030, against a no-AI path that grows 6.5% a year.
| Scenario | India GDP | US GDP | India cognitive unemployment | Export jobs vs no-AI path | Residents’ gain ÷ cognitive loss | US gain ÷ cognitive loss |
|---|---|---|---|---|---|---|
| Calculating… | ||||||
Set what you expect AI to be able to do, how widely it is adopted in India and by India’s clients, and how easily India adjusts. The model recalculates as you move each control; grey lines show the three scenarios for reference.
Calculating…
Solid: GDP. Dashed: income of Indian residents. Grey: the three scenarios.
Dotted: the normal rate of 8.5%. Grey: the three scenarios.
Red: cognitive workers. Blue: all other workers.
IT and business-services exports. Grey: the three scenarios.
The economy is a bundle of tasks. AI affects a share of cognitive tasks, is used on a share of their instances, and makes each AI-performed instance cheaper. Some AI use automates the task, moving it to capital; the rest assists a worker. New tasks for people can arrive as old ones are automated.
Workers belong to two groups: cognitive occupations (management, professional, technical, clerical and sales work) and all others. Cheaper cognitive tasks raise demand for the other group’s work, so its pay rises, while cognitive workers face lower demand. Their pay adjusts slowly, firms stop replacing departures and then lay off, and displaced workers search for jobs with a penalty when they switch occupation.
The India version adds an export island whose demand falls as the US scenario automates offshored work; displaced export workers keep only part of their earning power elsewhere. It splits income on automated tasks between Indian and foreign owners, and it weights workers by their share of labour income, which matters because Indian cognitive workers earn about 1.9 times other workers.
Calibration uses PLFS unemployment, Penn World Table labour share, Nasscom export and headcount figures, MoSPI national accounts and Anthropic’s India usage data. Several labour-market flows are assumptions because India has no matched flow survey; the report lists each one and tests the important ones.
What the model leaves out: earnings losses when a domestic cognitive worker moves into other work, underemployment in farm and casual work, demand spillovers in tech cities, the exchange rate, policy responses, robotics, and anything after 2030. On balance these omissions make the picture kinder to cognitive workers than it would be.
The model reproduces the original paper’s US results to within rounding, and the US figures shown beside each result come from the same code. The full report, with calibration tables, robustness checks and sources, is here; the original explorer is at anthropic.com/institute/econ-scenarios.