The BaseResearchWorld Bank
World Development Report 2026: The Promise of Artificial Intelligence
The World Bank's 2026 World Development Report assesses what AI means for low- and middle-income countries. Using an ILO-NASK exposure index, it estimates about 4.5% of existing jobs there are amenable to automation by generative AI, against 14.2% in high-income countries, and cites early falls in job postings for exposed white-collar work. Read from its Overview booklet.
Why it matters
The same thing, for different readers.
What it means for you
More jobs in poorer countries are amenable to AI complementing than automating, and most people there do manual work in small firms; early posting declines are in white-collar jobs.
For lawmakers
Exposure figures cover existing tasks, not realized job loss; the report urges building skills, infrastructure and institutions, and avoiding tax and regulatory rules that encourage replacing workers with AI.
For researchers
Applies the ILO-NASK exposure index to 137 countries; the jobs evidence rests on a few posting studies, and the report notes its estimates ignore new tasks AI could perform.
The question it raises
Who should share in AI's gains? The report warns benefits may flow to owners of capital and that poorly designed AI subsidies could leave taxpayers bearing the risk.
In their own words
Every quotation was found, word for word, in the text we saved from the source. That shows the words are there; it does not make them true, or prove the speaker meant them as quoted.
From the source’s own text.
“In low- and middle-income countries, about 4.5 percent of existing jobs are amenable to automation by generative AI—compared with 14.2 percent in high-income countries—while 16.2 percent of jobs are amenable to being complemented by AI (refer to figure O.6).”
Found word for word in the source
“Early evidence reveals that the launch of ChatGPT reduced the demand for jobs that are more amenable to automation by AI, although the impact is much smaller in low- and middle-income countries than in high-income countries.”
Found word for word in the source
“However, early evidence suggests that a growing share of the economic gains from AI is going to owners of technology and capital rather than to workers.”
Found word for word in the source
“But definitive evidence that AI is accelerating economic growth, generating more and better jobs, and improving service delivery at scale—the outcomes that shape people’s daily lives—is hard to find.”
Found word for word in the source
“These estimates might understate AI’s potential to complement workers because they are based on existing tasks and do not consider any new tasks that AI could perform.”
Found word for word in the source
How the finding was reached
In our words, from the source. Weigh the finding by how it was made.
- What was done
- A World Bank flagship report; this entry read its Overview booklet. The jobs shares use the ILO-NASK task-exposure index (the ILO working paper in this pass) for 137 countries. Other evidence in the Overview: World Bank Enterprise Survey on AI adoption in firms in India, Jordan, Kenya, Mexico, Nigeria, Thailand and the United States; Lightcast online job postings for more than 100 countries; a survey of governments; and studies of job postings after ChatGPT's launch, including in China and India. Growth figures are scenario estimates by the report team drawing on published studies.
- What it found
- About 4.5% of existing jobs in low- and middle-income countries are amenable to automation by generative AI, against 14.2% in high-income countries, and 16.2% of jobs there are amenable to being complemented by AI. Early evidence is that ChatGPT's launch reduced demand for jobs more amenable to automation, much less in low- and middle-income countries; the report cites falls in job postings in China, especially entry-level and highly educated workers, and in India for white-collar occupations most amenable to AI, especially entry-level. It calls these displacement effects narrow in scope but important because knowledge-intensive services are a growing share of middle-class jobs there, says AI is not yet replacing large numbers of jobs in most developing countries, and flags pressure on call-center and back-office outsourcing. In the most optimistic scenario AI raises potential growth in the 2020s from 1.2% to 3.6% a year in advanced economies and from 4.1% to 4.9% in developing countries.
- What it does not show
- The jobs shares score existing tasks, not realized job loss; the report says they might understate AI's potential to complement workers because new tasks are not counted, and that other studies using the median occupation as the cutoff report higher shares. The job-posting evidence comes from a few studies (China, India, ChatGPT's launch) and is described as early; the report itself says definitive evidence that AI is accelerating growth, generating more and better jobs, or improving services at scale is hard to find. Growth gains are scenario estimates, not observations. This entry read the Overview, not the nine chapters.
- Status
- A report
The record
- Questions
- What happens to work, income and the economy?
- People
- Indermit Gill (Wrote it, Senior Vice President and Chief Economist, The World Bank Group) · Gaurav Nayyar (Mentioned, Director, World Development Report 2026)
- Also covered by
- World Bank
- What we read
- We could read only part of the source.
- Found by
- A research pass: work, income and the economy, gathered October 13, 2026, via
research-work(tier 1: the source’s own channel or an official record) - Cite this
- World Bank, “World Development Report 2026: The Promise of Artificial Intelligence”, August 4, 2026. Entry in the Ctrl AI Base, https://ctrlai.com/base/world-bank-2026-08-04-wdr-2026-promise-of-ai
Our notesText: the 68-page Overview booklet of WDR 2026 (the file behind DOWNLOAD OVERVIEW on the report page), read for its jobs, growth, inequality and policy passages; the nine chapters were not read, and chapter 4 (jobs, productivity and growth) holds the detailed evidence, so access is partial. The saved text is pdftotext in reading order: fetch-text.sh in layout mode interleaves the two columns line by line and no quotation could be verified in it. The research pass saved the layout version, which has the same words. Date: the Overview gives none; 2026-08-04 is the release date in the World Bank press release (the report was released that day), and the booklet PDF's own file metadata says 1 Sep 2026. The Overview carries the Bank's usual statement that its findings do not necessarily reflect the views of the Bank, its Board of Executive Directors or governments. Quotations use the organization as speaker. The press release adds that 18.7% of jobs in high-income countries are amenable to complementing, which the Overview text does not give. The Foreword by Indermit Gill puts the same finding as less than a tenth of developing-economy jobs against more than a third in high-income economies, which differs from the Overview's 14.2%; the Overview's numbers are used. The Overview cites the NBER ChatGPT paper in this pass (endnote 82) for people using AI for practical advice, information and everyday tasks. The Overview also says US businesses pay under 5% tax on buying automation equipment or software, against an average of 25% in payroll and income taxes for hiring, under the heading on avoiding tax and regulatory policies that encourage replacing workers with AI: relevant to the AI-tax proposals elsewhere in this pass. The report's AI disclosure lists Claude models among tools used for coding and labelling data; the entry is not about Anthropic. The relation to the Bloomberg India entry rests on an excerpt-only entry (a one-minute clip description).