Stanford Graduate School of BusinessWorld Development Report 2026: The Promise of AI - How Will AI Impact Jobs and Productivity?
Stanford Graduate School of Business
Two illustrative AI examples 0:00
The speaker opens with two findings from the World Development Report. In Nigeria, among businesses that still run on handwritten records, about 20 percent already use AI, mainly through phone-based chatbots, showing how the technology can reach people who had almost no prior digital tools. In Kenya, entrepreneurs given an AI chatbot for business advice split sharply in outcomes: those already doing well saw revenues rise 15 percent, while struggling entrepreneurs saw performance fall 8 percent, suggesting AI can widen gaps rather than close them.
Three channels shaping jobs 3:32
The report frames AI's effect on jobs and productivity through three channels. First, AI as an input into work, which can replace tasks or make jobs more productive and in demand, with ambiguous net effects. Second, AI as an output, since low and middle income countries also participate in producing AI, creating jobs and productivity gains. Third, AI reshaping economies more broadly by changing costs and the location of production, with wider ripple effects on employment.
Current job exposure is limited 5:33
Today, most people in low and middle income countries work in jobs, like smallholder agriculture, retail, and personal services, that are not directly exposed to AI, so immediate job impacts and productivity gains look limited, offering some reassurance. The Philippines and India are exceptions due to their knowledge-intensive export services. But the sectors creating the most new jobs tend to be more exposed to AI, meaning the trajectory of employment could shift substantially as the technology advances.
Surveying AI adoption by firms 7:30
To fill a data gap on how firms in low and middle income countries use AI, the team surveyed over 4,000 formal businesses with five or more employees across seven economies, asking about usage, purpose, expectations, and technology sources, with all data and code made public.
Rapid but uneven adoption rates 9:00
By late 2025, 27 percent of surveyed businesses used some form of AI, a faster pace than earlier technologies like email or websites. The United States leads, but countries like Kenya and Nigeria are approaching similar adoption rates for basic AI tools such as chatbots. Adoption gaps widen for more advanced uses like robotic process automation or AI agents, where developing economies lag further behind.
Firm size and sophistication matter 11:31
Larger firms are more likely to adopt AI and to use it in sophisticated ways, such as coding agents, process automation, or customized solutions, a pattern seen in both the US and developing economies, though sophistication accelerates especially once US firms pass around 50 employees. Businesses with stronger digital readiness, management practices, and prior productivity are also more likely to adopt and use AI effectively, showing that existing capabilities shape who benefits.
Main barriers to adoption 14:31
When asked about obstacles, businesses most often cite a lack of knowledge about how to use AI productively, followed by cost concerns, especially in developing countries, and worries about privacy and security.
Policy framework: adopt, adapt, advance 15:30
For policymakers, the report's framework distinguishes adoption, adapting ready-made solutions, and advancing more sophisticated use. For most firms in low and middle income countries, adoption of near-ready solutions matters most, which depends on analog foundations like internet infrastructure and a dynamic business environment that rewards investment and well-performing firms.
Adopt, Adapt, Advance Framework 17:00
Helping businesses use AI well means thinking in three stages. Adoption requires basic literacy and numeracy so businesses can even ask the right questions, plus matchmaking and de-risking so viable tools reach firms. Adaptation, more relevant for medium and larger firms and startups building AI solutions, needs digital foundations like usable data, compute access, and supportive entrepreneurial ecosystems, along with deeper skills in building and adjusting models. Advancement, relevant to only a few firms in a few countries, depends on frontier infrastructure, research funding, international talent, and careful competition policy.
Report's Three Main Takeaways 20:00
Short-term AI impacts on low and middle income countries remain limited, but this is no reason for complacency since adoption and technology will shift quickly. Most firms are only adopters, with adaptation and advancement confined to a few, so policy must distinguish these roles. The core challenge is spreading adoption of relevant technology widely enough that it builds bridges to job creation and long-term growth.
Lack of Data and Knowledge Gaps 22:01
A second speaker stresses that writing this report was hard because there is so little existing evidence, and the field is moving fast. He suggests watching firm size distributions as an early signal, since smaller firms often act like a canary in the coal mine. He also warns that knowledge gaps exist not just among entrepreneurs but among policymakers deciding national investment priorities, which could slow AI's benefits in developing countries.
Small Business AI Adoption Data 26:01
Using data from very small US, UK, and Canadian businesses, he shows that while 70 to 75 percent claim to use AI, far fewer actually pay for it: under 10 percent pay even $10 for two months, and under 1 percent pay $100. Incumbent large firms like Oracle and Amazon are cutting tens of thousands of jobs mainly to reduce costs, but new small and young businesses are expected to create unforeseen opportunities, likely shrinking average firm size while increasing the number of entrepreneurs.
Sector and Firm Size Patterns 28:31
Information sector firms lead AI adoption while agriculture lags furthest behind, even in the US, hinting at missed potential for developing countries where agriculture matters more. Adoption by firm size follows a U shape, with one worker firms adopting more than two or three worker firms, suggesting constrained entrepreneurs use AI to expand tasks cheaply instead of hiring.
Transformative Entrepreneurs and Outcomes 32:03
Not everyone is suited to entrepreneurship, and most new businesses, like the surge during Covid, form out of necessity rather than vision. The few transformative entrepreneurs who aim to grow are far more likely to adopt AI, making it a potential gamechanger where such entrepreneurship is scarce. Surveys again show privacy concerns and lack of knowledge as top adoption barriers, a gap he notes is especially stark among policymakers. Among small businesses, AI-linked hiring already outpaces firing, consistent with creative destruction theory.
Revenue Productivity Among Small Businesses 34:00
Revenue growth among small businesses outpaces revenue decline, meaning revenue per worker is rising without major job losses. That said, developing countries face a weaker outlook, hampered by basic infrastructure gaps like frequent power outages, especially in Africa, where entrepreneurs described having to restart model training from scratch every time electricity fails. The speaker argues that not adopting AI carries a large opportunity cost for developing economies, since frontier countries are moving extremely fast, with growth projections as high as ten to fifteen percent.
Building an AI Monitoring Index 38:02
Working with the Knowledge Lab, researchers built an index using academic papers, patents, grants, GitHub developer activity, data center information, and business registry data to rank countries on AI readiness. The United States ranks first, followed by China and the United Kingdom. Comparing national AI strategies shows different emphases: the US stresses international security, China focuses on R&D and social welfare, and Europe leans toward regulation in a cautious, safeguarding way.
Matching Strategies to Real Needs 42:00
Cross-referencing these strategies with World Bank governance data reveals that countries with more STEM graduates still talk more about education, while countries with weak government effectiveness or rule of law talk more about regulating AI. Plotting countries by infrastructure gaps against how much they discuss those gaps shows some, like Germany and China, discussing power issues despite reliable electricity, while others face outages but say little about it, a mismatch the speaker flags as worth monitoring.
Can AI Lift Small Businesses 45:00
Asked whether AI can be a game changer for the long tail of unproductive small businesses in developing economies, one speaker notes that business entry rates are high but entrants are often no more productive than the firms exiting, breaking the usual engine of creative destruction. Early evidence suggests AI helps entrepreneurs who hire technical or white-collar staff early, which strongly predicts future growth, so AI could act as an enabler alongside merit-based competition.
Entry, Scaling, and Reallocation Problems 47:30
Another speaker adds that the deeper issue is failure of productive entry and failure of firms to scale, illustrated by Ghana follow-up surveys showing businesses unchanged a decade later. Finance and employment often fail to flow toward the most productive firms, creating a reallocation problem. Business development and technical assistance can help, as seen in Kenya, but results depend heavily on local policy conditions, since such support does not work equally well everywhere.
Local Context And National AI Strategies 51:00
A Stanford alum working with startups through Y Combinator asks how national AI strategies address local context, since many countries still lack clarity on what should be trained into models versus imposed from outside. In response, the speaker clarifies that revenue growth outpacing employment growth suggests AI adoption is linked to productivity gains. He points to Africa, where 2,000 indigenous languages mean AI tools could bridge people to public services like education and healthcare, provided governments make their data available for training and buy into using their archives this way.
Codifying Local Knowledge For Africa 53:30
Elvin describes working with a business development provider in Ethiopia, where outcomes depended on details like which valley a farmer worked in or when they planted and harvested. Much of that knowledge lives only in people's heads, uncodified, so building usable datasets from local languages and conditions is a major challenge, one that initiatives from Google and others are starting to address.
Plug And Play Tools For Small Firms 54:30
On adapting AI for small and micro enterprises, Elvin says this remains an experimental area. Many businesses track production and inventory poorly, and AI could help by scanning receipts or using cameras to log activity automatically. The real opportunity is building simple, ready-to-use plug-and-play tools for common business types like retail, since many such businesses resemble one another closely.
Job Displacement In Service Sectors 56:00
A Santa Clara University professor raises concern that AI risk, though statistically lower in low and middle income countries, threatens exactly the higher-income service jobs women and young people were moving into, such as nursing and education, while informality may rise, particularly affecting women in regions like the Middle East and North Africa.
Electricity And Supporting Small Business 59:31
Other questions cover whether the World Bank addresses electricity scarcity for AI computing, noting China's shift of data centers to desert regions with solar power, and how to help small businesses scale beyond individual cities through training and policy support.
Reframing The Informality Debate 1:00:02
The response argues that developing countries were not thriving before AI, so the real question is not whether to slow the technology but how to use it as a stepping stone to restructure economies stuck in informality. Informality persists because formal firms fail to scale, not because informality itself is large; AI offers a rare chance to push transformative entrepreneurship that displaces inefficient and informal firms.
World Bank Energy And Digital Programs 1:02:02
On electricity, the World Bank works with governments and the IFC to build infrastructure but cannot solve the problem alone, requiring private capital mobilization. Shamik clarifies the Mission 300 program aims to bring energy access to 300 million people in Africa by 2030, focused on basic reliable access rather than compute infrastructure. A related digital initiative aims to connect 600 million people to broadband and 400 million to basic digital services.
Closing Reflection On Analog Complements 1:03:32
The session closes with the observation that as AI technology advances, the analog complements, meaning the basic knowledge, awareness, and information people need to use these tools effectively, become more important, not less. Both presenters stressed that many people simply do not know what AI can do for them or how to get good returns from it, and this gap will likely shape how quickly AI gets adopted for productive use.
AI-generated summary. It can be wrong or incomplete - check anything that matters against the original.
