Rethinking Automation Risk in Ethiopia’s Labor Market

Ten years ago, a widely cited study from the Oxford Martin School, using World Bank data, estimated that as much as 85 percent of jobs in Ethiopia were at risk of automation, shaping early debates on automation risk in Ethiopia and the broader future of work in developing countries. The finding helped popularize concerns about “premature deindustrialization,” where job automation in low-income countries could outpace industrial growth before it fully matures. A decade later, with rapid advances in Generative AI (GenAI), AI-driven task automation, and digital technologies, alongside new research on task-based automation vs occupation-based automation, the key question is whether Ethiopia’s exposure to automation has intensified or declined.

Recent evidence suggests the answer is more nuanced than the original headline implies. Early studies like The Future of Employment relied on occupation-level analysis, meaning they assessed entire jobs as automatable or not. Recent research has shifted toward task-level analysis and occupational exposure metrics, which yield lower, more realistic estimates. For example, a 2025 analysis, based on financial and central bank data, finds that less than 10 percent of core job tasks can currently be fully automated by AI systems. This distinction matters because most jobs consist of a bundle of tasks, many of which remain resistant to automation due to social interaction, non-routine cognitive adaptability, or physical context.

Structural features of Ethiopia’s economy further reduce immediate automation risk. Much of the labor force is concentrated in agriculture and informal services, sectors that are difficult to automate due to low capital intensity and high reliance on manual and context-specific work. At the same time, industrialization efforts have created factory jobs that remain labor-intensive. Experimental evidence from Ethiopian industrial parks shows that while factory employment can increase short-term earnings, these jobs remain relatively low-productivity and are not easily substituted by advanced automation technologies in the near term.

However, this does not mean the risk has disappeared. Instead, it has evolved. New research on developing labor markets emphasizes “routine-biased technological change,” where tasks that are repetitive and predictable are increasingly automated, while non-routine work becomes more valuable. This creates uneven exposure within countries. Urban, educated, and formal-sector workers may face higher AI exposure than rural populations, reversing earlier assumptions that automation primarily threatens low-skill labor.

Another emerging concern is the difficulty of transitioning between jobs. Although studies in Ethiopia don’t specifically cover this, we can look to other countries for insight.  A 2026 study on Egypt’s job market found that only about a quarter of workers in high-risk roles could easily move into safer occupations without significant retraining. This suggests that even if fewer jobs disappear outright, adjustment costs could be substantial, especially in countries with limited reskilling infrastructure.

There is also growing evidence that AI’s primary economic effect may be on wages rather than employment. Recent empirical work shows that AI exposure is associated with slower wage growth even when employment levels remain stable. For Ethiopia, where wages are already low, and labor markets are tight, this dynamic could be more consequential than outright job displacement.

Taken together, the best available evidence indicates that the original 85 percent estimate overstated the short-term risk by treating jobs as fully automatable units. Advances in AI have increased the range of tasks that machines can perform, but empirical data shows that large-scale job elimination has not materialized. Instead, the risk profile has shifted toward gradual task transformation, uneven exposure across sectors, and potential pressure on wages.

The real challenge for Ethiopia is no longer a sudden wave of automation-driven unemployment. It is the slower, more complex process of adapting the workforce to a changing structure of work. Education systems, vocational training, and industrial policy will determine whether technology complements labor or quietly erodes its value over time.