A recent study by Anthropic finds that nearly 70% of tasks for programmers, customer service reps, and data entry operators are already automated, making these roles the most exposed to displacement by large language models (LLMs). The research introduces an "observed exposure" metric, showing a significant gap between AI's theoretical capabilities and its current real-world use in professions.
Other fields with high observed exposure include medical recordkeeping, marketing, sales, and financial analysis, raising concerns for service-dependent economies like India. While LLMs have high theoretical potential in fields like computer and mathematics, their actual task coverage is currently much lower, and many physical or complex interpersonal tasks remain beyond AI's reach.
Nearly 70% of tasks performed by programmers, customer service representatives, and data entry operators are already automated, according to a recent study by Anthropic. The findings suggest that these roles face the highest risk of being displaced due to the growing use of large language models (LLMs).
âProgrammers are at the top, with 75% coverage (automation), followed by Customer Service Representatives⦠Finally, Data Entry Keyers, whose primary task of reading source documents and entering data sees significant automation, are 67% covered,â the study said.
Software developers largely write and update code, while for call centre agents the job includes taking orders and handling customer complaints; data entry specialists read documents and enter information into systems.
Among the top 10 professions experiencing high levels of what Anthropic calls âobserved exposureâ to LLMs are medical recordkeeping, marketing, sales, and financial and investment analysis. This reinforces concerns for economies like India, where growth and employment remain heavily dependent on the services sector.
Anthropic devised this observed exposure metric to estimate the extent to which tasks that LLMs theoretically seem to be able to automate are actually being used in real professional settings. Unlike theoretical capability, which covers a broad range of possible AI applications, observed exposure focuses on where AI is already affecting work. By tracking this gap between potential and real-world use, the measure helps capture the emerging economic and labour market impact of AI.
Also, while LLMs could theoretically affect the majority of tasks in occupations such as `computer and mathematicsâ (94%) and `office and administrative rolesâ (90%), actual usage remains much lower.
Data from the Anthropic Economic Index indicates that AI tools like Claude currently cover only about 33% of tasks in the computer and mathematics category. At the same time, many tasks, particularly those which involve physical labour, such as agricultural work, or activities like representing clients in court, remain largely beyond AIâs reach.
âProgrammers are at the top, with 75% coverage (automation), followed by Customer Service Representatives⦠Finally, Data Entry Keyers, whose primary task of reading source documents and entering data sees significant automation, are 67% covered,â the study said.
Software developers largely write and update code, while for call centre agents the job includes taking orders and handling customer complaints; data entry specialists read documents and enter information into systems.
Among the top 10 professions experiencing high levels of what Anthropic calls âobserved exposureâ to LLMs are medical recordkeeping, marketing, sales, and financial and investment analysis. This reinforces concerns for economies like India, where growth and employment remain heavily dependent on the services sector.
Anthropic devised this observed exposure metric to estimate the extent to which tasks that LLMs theoretically seem to be able to automate are actually being used in real professional settings. Unlike theoretical capability, which covers a broad range of possible AI applications, observed exposure focuses on where AI is already affecting work. By tracking this gap between potential and real-world use, the measure helps capture the emerging economic and labour market impact of AI.
Also, while LLMs could theoretically affect the majority of tasks in occupations such as `computer and mathematicsâ (94%) and `office and administrative rolesâ (90%), actual usage remains much lower.
Data from the Anthropic Economic Index indicates that AI tools like Claude currently cover only about 33% of tasks in the computer and mathematics category. At the same time, many tasks, particularly those which involve physical labour, such as agricultural work, or activities like representing clients in court, remain largely beyond AIâs reach.
Meridiano
