Every wave of automation has disrupted the labour market. The industrial revolution displaced agricultural and artisanal work. Similarly, computerisation in the 1980s and 1990s eliminated large numbers of clerical and administrative roles. Likewise, the robotics revolution in manufacturing displaced production-line workers. In each case, economists pointed out — correctly — that new jobs emerged as old ones disappeared. They also pointed out that the long-run effect on employment levels was broadly neutral. In addition, they pointed out that the long-run effect on living standards was positive.
However, the distribution of these transitions has never been neutral. The people whose jobs automation replaced did not automatically transition to the new jobs that emerged. They faced periods of unemployment, skill mismatches, geographic displacement and income loss that could last years or decades. In other words, the long-run aggregate story coexists with a short-run individual story that is often one of genuine hardship.
The current AI wave has a distinctive feature that distinguishes it from its predecessors. It disproportionately affects high-skill, high-wage, white-collar work. This inverts the pattern of previous automation waves. As a result, it creates a different set of distributional consequences. These consequences are politically and socially complex in ways that the displacement of manufacturing workers was not. Even so, that displacement was morally no less serious.
Why This Wave Is Different
Eloundou et al. (2023) analysed large language model capabilities and the US workforce. They found that GPT-level AI significantly affects at least 10% of tasks for approximately 80% of the US workforce. In addition, it affects more than 50% of tasks for 19% of the workforce. Crucially, the analysis found that “higher wages and higher educational attainment generally have higher exposure to GPT capabilities”. This is a direct inversion of earlier automation patterns.
Goldman Sachs (2023) identified the occupational categories most exposed to generative AI automation. These included administrative and office support work (46% of tasks exposed) and legal work (44%). They also included architecture and engineering (37%) and business and financial operations (35%). In contrast, construction and maintenance occupations had just 6% of tasks exposed. In fact, the world in which graduate-level professionals feel more job insecurity than skilled tradespeople is arriving.
The Roles Facing the Most Direct Disruption
The World Economic Forum’s Future of Jobs Report 2025 identified the fastest-declining roles by 2030. These are postal clerks and sorting-office workers, bank tellers and related staff, and data-entry clerks. They also include cashiers and ticket clerks, administrative and executive secretaries, and accountants and auditors (WEF, 2025). Frey and Osborne (2017) found that telemarketers, tax preparers, insurance underwriters and data-entry keyers all had automation probabilities exceeding 99%.
In legal services, the impact is well documented. AI tools now perform document review, contract analysis, legal research and due diligence. They do so at speeds and costs that are dramatically lower than equivalent human effort. The number of paralegal and legal assistant roles in large firms has already begun to decline. In addition, AI is disrupting the entry-level pipeline for qualified solicitors and barristers. This pipeline has historically run through years of junior document work. The Law Society of England and Wales has estimated the scale of this impact (Law Society, 2023). Specifically, it suggests AI could affect up to 67% of tasks that UK lawyers perform.
In financial services, AI is automating tasks at scale. These tasks include financial analysis, portfolio management, compliance monitoring, fraud detection and customer service. BlackRock, Vanguard and other major asset managers have reduced headcount in analyst and portfolio construction roles. In these roles, AI tools now perform the equivalent function more cheaply and with fewer errors. The FCA’s ongoing review of AI use in UK financial services reflects the regulatory implications. These implications matter because AI already pervades the sector’s core operations.
In journalism, marketing and content creation, generative AI has already automated significant volumes of routine content production. For example, AI systems at major publishers and retailers now routinely generate sports match reports and earnings summaries. They also generate weather forecasts and product descriptions. The roles that are contracting are not those of senior investigative journalists or creative directors. Instead, they are those of junior content creators, sub-editors and production staff. AI can replicate their work at a fraction of the cost.
The Early-Career Problem
A particular concern in the academic and policy literature is the impact of AI on early-career professionals. The most automatable tasks tend to be the foundational tasks that junior professionals have traditionally used to develop their skills. For example, junior lawyers use document review and junior analysts use financial modelling. Similarly, junior software engineers use code review, and junior researchers use literature review. If AI increasingly performs these tasks, the question of how junior professionals develop their skills is genuinely urgent. In addition, whether the pipeline from novice to expert remains functional is equally urgent.
Brynjolfsson, Li and Raymond (2025) found that AI assistance raised customer service workers’ productivity by an average of 14%. Specifically, the largest gains — 34% — came among novice workers. This is encouraging for individual workers who have access to AI tools. However, a novice worker with AI support begins to produce output that approaches an unaided expert’s. This reduces the demand for junior roles and potentially hollows out the middle of professional hierarchies.
What Should Individuals Do?
The honest answer is that no individual response entirely insulates anyone from the labour market consequences of AI-driven automation. The disruption is systemic. As a result, it requires systemic — policy, institutional, collective — responses alongside individual ones.
However, the research consistently identifies several responses that are genuinely helpful. Developing skills in working alongside AI tools has consistently correlated with wage premiums and employment resilience. These skills mean using AI tools as productivity amplifiers rather than substitutes. PwC’s AI Jobs Barometer (2024) found a 25% wage premium for skills linked to AI. It also found that industries most exposed to AI were seeing 4.8 times higher productivity growth than less-exposed industries. As a result, the workers gaining in this environment are those who leverage AI to do more and better work. They are not those competing with AI by trying to do the same work humans can do without assistance.
In addition, developing genuinely human skills is the other consistent recommendation. These are the relational, ethical, creative and embodied capabilities that AI cannot replicate. Examples include the ability to navigate complex human relationships and to exercise ethical judgement in ambiguous situations. They also include the ability to communicate persuasively in person and to provide genuine care and connection. As a result, they will command a premium in a world where AI increasingly automates routine cognitive work. These are not vague reassurances. In fact, the same research that identifies which tasks are and are not automatable supports them.
References
- Brynjolfsson, E., Li, D. and Raymond, L.R. (2025) ‘Generative AI at work’, Quarterly Journal of Economics, 140(2).
- Eloundou, T. et al. (2023) ‘GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models’, arXiv preprint, arXiv:2303.10130.
- Frey, C.B. and Osborne, M.A. (2017) ‘The future of employment: How susceptible are jobs to computerisation?’, Technological Forecasting and Social Change, 114, pp. 254–280.
- Goldman Sachs (2023) The Potentially Large Effects of Artificial Intelligence on Economic Growth. New York: Goldman Sachs Economics Research.
- Law Society of England and Wales (2023) The Future of Legal Services: AI and the Legal Profession. London: The Law Society.
- PwC (2024) AI Jobs Barometer 2024. London: PricewaterhouseCoopers.
- World Economic Forum (2025) Future of Jobs Report 2025. Geneva: World Economic Forum.