Future of work: Which type of labor is AI most likely to displace?

A new academic analysis challenges widely held assumptions about which sectors are most vulnerable to artificial intelligence (AI), proposing a fundamental rethinking of labor policy in the digital era. In his latest study, philosopher and policy theorist Yotam Harel unpacks the deepening disruption caused by artificial intelligence (AI) in global labor markets and explores how a structurally just response could be architected before unemployment reaches irreversible levels.

The study, titled "AI, Mental and Physical Labor, and a Just Policy Framework", is published in AI & Society. The author argues that existing policy discourses have underestimated AI’s capacity to automate and eventually marginalize mental labor, historically seen as the most stable domain of human contribution. The study concludes that while both are at risk, mental labor, owing to the generality and scalability of AI, faces a more immediate threat. 

Which type of labor is AI most likely to displace?

The study asserts that the defining features of mental labor make it particularly vulnerable. Mental labor includes occupations involving symbolic reasoning, language generation, pattern recognition, decision-making, and other cognitive tasks often performed by professionals such as writers, analysts, designers, accountants, and programmers. Unlike physical labor, which typically involves complex environmental interaction and embodied dexterity, mental labor can often be encoded, scaled, and replicated via software.

AI's two defining traits, generality and near-zero marginal cost, allow it to be deployed across a vast range of intellectual tasks without requiring significant additional resources. As a result, roles once reserved for university-educated knowledge workers are increasingly automated by language models, predictive algorithms, and expert systems. The study outlines how this trend will likely accelerate, leading to mass displacement in fields once considered immune to automation.

By contrast, physical labor remains comparatively safe in the short term. This category includes jobs such as cleaning, caregiving, food preparation, and construction. Despite efforts to develop robotic systems capable of performing these roles, the physical world presents far more unpredictable and complex variables than data-rich mental tasks. Therefore, many physical labor jobs will persist, though often in low-paid, precarious forms.

What societal risks arise from unequal AI-driven displacement?

The study situates its analysis in a broader socio-political context, warning of significant risks if the AI-induced restructuring of labor continues without state intervention. A major consequence would be a vast underclass of unemployed or underemployed mental laborers who were previously central to economic productivity and civic life. These individuals may not possess the physical aptitude or opportunity to shift into remaining physical labor roles, creating long-term structural unemployment.

As AI tools replace professionals in white-collar fields, wealth creation is likely to concentrate in the hands of a few firms that own or control AI infrastructure. This consolidation risks deepening existing economic inequalities, potentially leading to widespread social unrest, loss of democratic legitimacy, and civic alienation. Moreover, because physical labor remains undervalued and underprotected in many economies, even a shift toward increased demand in these sectors may not translate to greater stability or prosperity for displaced workers.

The study also warns that even physical labor’s relative resilience may be temporary. As robotics and embodied AI systems advance, tasks involving physical dexterity, spatial awareness, and mobility may also become subject to automation. In this scenario, the entire labor system faces disruption, prompting the urgent need for a coherent ethical and policy response.

What policy framework could address the challenges ahead?

The study provides a forward-looking policy proposal built on principles of justice, equity, and systemic foresight. At its core is a job guarantee program, wherein the state ensures employment opportunities for all who seek work. These could be concentrated in public goods sectors like environmental restoration, caregiving, community health, education support, and infrastructure maintenance, areas traditionally underserved by market forces but crucial to human well-being.

Another pillar of the framework is Universal Basic Income (UBI). Recognizing that AI’s productivity does not require full employment, UBI would provide a stable financial floor for all citizens, decoupling livelihood from traditional employment. This would help avoid the moral hazard and inefficiency of mass unemployment while preserving dignity and freedom of choice for workers navigating a transformed labor market.

The study also advocates for targeted retraining and reskilling programs, especially for mental laborers seeking to transition into hybrid roles involving human oversight of AI systems, or into physical labor sectors where demand remains. These programs must be inclusive, long-term, and publicly funded to ensure access and avoid reinforcing existing educational or gender disparities.

To fund such programs, the author proposes AI taxation, an economic instrument designed to capture a portion of the gains generated by AI deployments. This could involve taxing firms based on the scale of AI-based productivity they generate or reducing the corporate incentives for replacing human labor with automated systems.