When Human Jobs Vanish: AI Shaking the Human Domain and the Entry Point of Work (updated September 19, 2026)

Companies are deciding not to outsource work to humans if it can be processed by AI—a policy that has become increasingly apparent starting in 2025. The targets are not factory operations, but desk jobs such as creating educational materials, answering inquiries, and summarizing documents.

What is happening is not so much the wholesale disappearance of occupations, but rather a change in design: which tasks to hand over to AI, which decisions to leave to humans, and whom to hire for the work. When the design changes, both the nature of the work for current employees and the initial tasks assigned to those entering the workforce will change.

History is long with tools reducing work, and some professions, like telephone operators, have vanished. Yet, the current anxiety is particularly strong because AI has begun to take over tasks—writing, responding, and summarizing—that were previously considered the exclusive domain of humans.

1. Corporate AI Implementation and Shifting Employment Forms

One example that clearly demonstrated such a policy is Duolingo. The company used generative AI to develop new language courses for 148, dramatically accelerating the speed of content production. Along with this, CEO Luis von Ahn indicated an internal policy to gradually stop using contractors for tasks that AI can handle. Highlighting the results of AI-driven scaling, he noted that while the first 100 course took 12 years to develop, nearly 150 courses were built in about one year after the introduction of AI (according to TechCrunch reporting).

In response to this policy, concerns about workforce reductions erupted externally. Regarding this, von Ahn explained that the policy was misunderstood, stating that no full-time employees were laid off and that the number of contractors naturally fluctuates based on demand (according to an interview with the CEO). However, this served as a concrete example of a decision to stop outsourcing to humans in areas where AI can substitute.

Similar moves are not limited to tech companies like Duolingo. The 2025 fiscal year reports of BT Group and Klarna also list digitalization of operations and technological innovation as strategic priorities (BT Group Annual Report, Klarna Annual Report). However, these reports do not show actual results of employment replacement by AI. They are evidence that companies are linking operational design with digitalization and technological innovation.

2. From Routine to Open-Ended Intellectual Work

The reason anxiety over AI replacing jobs is more acute than during past automation is that the target of automation has expanded from "routine tasks" to "open-ended intellectual work."

In the case of the mechanization of telephone operators in the United States at the beginning of the 20 century, automation at the time replaced specific routine tasks. While many jobs disappeared as a result, analysis suggests that this employment shifted to other roles, such as mid-skill clerical work or low-skill service work, and did not lead to a decrease in the total number of jobs for the next generation (according to NBER research). Furthermore, in the U.S. from 1990 to 2024, the automation of routine cognitive tasks through computerization progressed in fields such as accounting, bookkeeping, and tax preparation; however, there are cases where employment in these occupations actually increased, doubling from 600,000 to 1.2 million (according to a report from the National Academies).

However, the impact brought by generative AI is qualitatively different. The singularity of generative AI lies in the fact that the possibility of automation has extended to domains such as writing, responding, and summarizing, which were previously regarded as "human-specific open-ended intellectual work."

Past automation replaced rule-based routine work. Generative AI is now capable of handling parts of the intellectual work that humans have previously managed, such as expression and analysis. Consequently, a risk has emerged where advanced cognitive abilities, once thought to be the sole possession of humans, are provided as functions that machines can substitute.

3. The Metric of Scarcity Value in Expertise

The market value of intellectual labor changes significantly depending on whether that expertise is necessary to achieve a goal and whether only a few people possess it.

Generally, the market value of a task is guaranteed by two conditions: that the expertise is necessary for achieving the goal, and that only a small number of humans possess that ability. For example, the reason air traffic controllers' wages are significantly higher than those of crosswalk guards is that while both share the commonality of making quick decisions to avoid collisions, the former requires rare expertise—advanced training and certification—that takes time and cost to acquire (according to the report from the National Academies).

One of the critical risks brought by AI is the erosion of this scarcity value of expertise. As AI provides specialized capabilities as a low-cost alternative, expertise that previously maintained a high market value may become commoditized, and the scarcity of skills held by humans may be lost.

This is the same mechanism as the case where the knowledge of back alleys acquired by London taxi drivers lost value and earnings decreased due to GPS, or the case where a taxi dispatch program in Yokohama erased the gap in route selection between skilled drivers and novices (according to the same report). In intellectual labor as well, if AI can provide professional answers and advanced analysis at a low cost, human expertise may cease to be a scarce resource, risking a decline in wages and treatment.

4. Fragmented Impacts and the Precariousness of "Entry Points"

The impact of AI is not uniform; it creates divisions based on region, occupation, and career stage. A particularly concerning risk is the narrowing of career entry points for young workers.

According to a joint study by the ILO and the World Bank, it is pointed out that in developing countries, the impact of AI on mid-skill jobs, such as clerical work, risks narrowing the pathways to stable employment for youth and women. It is stated that regions with insufficient digital infrastructure compared to developed nations may experience a disproportionate shock—where existing jobs change—before they can reap the benefits of productivity gains (Joint research by the ILO and World Bank).

On the other hand, in markets such as the ASEAN region, while it is estimated that approximately 80 million people will be affected by AI, no evidence of large-scale unemployment has been found at this time (ILO ASEAN Survey). Furthermore, it is pointed out that high AI exposure should be viewed as a leading indicator of the transformation of job content, rather than immediately signifying the disappearance of employment (ILO Research Brief).

From this point forward, rather than results directly measured in the materials, this is a inference regarding skill formation. Much professional expertise is built by repeatedly performing basic tasks and gaining experience. If the basic tasks that serve as the entry point to intellectual labor are replaced by AI, the training opportunities for young workers to acquire expertise may decrease. The substitution of fundamental work required for gaining experience leads to the risk of undermining the pathways for cultivating the next generation of professionals.

5. Reorganized Work and Questions for Humanity

When considering employment in the AI era, the changes cannot be fully captured simply by whether jobs disappear or not. While unemployment rates in developed nations are at historic lows, it has also been pointed out that the value of specific expertise may decrease due to technology (National Academies Report).

According to forecasts by the World Economic Forum (WEF), 170 million new jobs will be created by 2030, while 92 million will be lost, resulting in a net increase of 78 million. However, this forecast is not based on the impact of AI alone, but on broad corporate outlooks including demographics, decarbonization, and geopolitical risks (World Economic Forum Announcement).

Looking at employment as a whole, the impact of AI may manifest not as the wholesale disappearance of jobs, but as the reorganization of work. Whether the result leads to worker well-being or expands inequality depends not only on the capabilities of AI but also on the institutional design on the human side.

One of the critical conditions that will determine the outcome of the future labor market is the decision-making regarding who enjoys the profits from efficiency and which judgments are intentionally left to humans. Policy choices, such as investment in skill development and the construction of social protection systems, as well as business process design by companies, will influence the value of work in the AI era.

This discussion has utilized Duolingo's policies, the annual reports of BT Group and Klarna, past automation cases in the U.S., and international materials from the ILO and WEF. Based on these materials alone, it is not possible to judge the employment changes by occupation in Japan, the impact on wages at individual companies, or the actual extent to which skill formation for young workers has changed.