Gauging disruption: AI studies highlight labour-market inequality

Anthropic, one of the biggest disruptors in artificial intelligence (AI), recently released two studies: One tracking usage patterns in large language models (LLMs) and another examining its early effects on the labour market. Although only data from Claude has been used for the analysis, the studies do offer a snapshot of where the technology stands today. LLMs are used more intensely in high-income countries, and for a relatively small set of specialised tasks. The central finding across both reports, however, points to a diversification in use cases and, correspondingly, a marginal decline in the average economic value of the work done. The study notes that early adopters favour specific high-value uses like coding while later adopters take on a much wider range of tasks, including personal queries. Also, global adoption is uneven, with the top 20 countries accounting for 48 per cent of per capita usage.

 

The differences in adoption can further deepen inequalities in the labour market, amplifying the effects of skill-biased technological change. In the second study, which introduces a measure for understanding the labour-market effects of AI, occupations such as those of computer programmers, customer-service representatives, and financial analysts are cited as among “the most exposed” to AI, to the extent that their tasks are theoretically feasible with LLMs. Thus, in sectors such as computer science, finance, legal service, and management, LLMs could, in principle, handle a majority of the tasks. But in practice, only a fraction of this potential is being realised. Early labour-market signals are beginning to emerge, highlighting no systematic increase in unemployment. Instead, hiring younger workers, particularly in entry-level positions, has slowed in exposed occupations.

 

In Claude usage, India ranks 98th among 116 countries. Despite the low overall adoption of advanced AI tools, Indian users are disproportionately deploying AI for coding and code debugging, design work, academic assistance, web development, and software development. Clearly, AI is being used as an instrument for enhancing productivity and employability. Nevertheless, the fact that experienced users are deriving greater gains through learning-by-doing is deepening the skill divide in the labour market. It also exposes a deeper vulnerability. The very sectors where Indian workers are highly concentrated, including information-technology (IT) services, back-office operations, and routine cognitive work, are among the most exposed to AI-driven automation. Early market signals are already visible. IT stocks are facing pressure and analysts are beginning to factor in the possibility that a meaningful share of revenues could be eroded as AI tools take over tasks that once required human teams.

 

If entry-level roles shrink because of AI, the pathway for skill accumulation itself may be disrupted. The most immediate priority, therefore, is to augment capability. Since AI is a skill multiplier, workers who can effectively integrate it into their workflows are already more productive and more resilient. This makes large-scale upskilling essential, not just in coding but in problem-solving and the ability to collaborate with AI systems. Thus, education and training systems need to quickly adapt. In this regard, the Central Board of Secondary Education’s launch of AI and computational thinking in the curriculum at school level is a timely intervention. AI-related courses need to be embedded in schools and institutions of higher education.