Mapping the Future of Work: AI’s Task-Level Transformation
Memo: February 11, 2025
Background
AI continues to advance rapidly and there is growing debate about their impact on the workforce and the broader economy. Traditional approaches to studying AI’s potential impact have relied on predictive models, controlled productivity studies, or periodic surveys, each with inherent limitations in capturing real-world use cases.
Recent empirical research from Anthropic improves upon previous studies by using Clio — a privacy-preserving analytical framework — to systematically measure AI usage across various work tasks, as defined by the U.S. Department of Labor’s O*NET Database. By mapping real-world conversations to occupational tasks, Anthropic’s research provides a dynamic and empirical snapshot of AI’s impact. Anthropic’s findings serve as a leading indicator of AI’s evolving economic impact.
Analysis
Key Findings and Their Implications
Concentration in Specific Sectors: Claude usage is predominantly found in computer-related tasks, writing, and analytical functions. These areas account for nearly half of all AI interactions.
Depth of AI Integration: Although a significant portion (∼36%) of occupations employ Claude for at least 25% of their tasks, only about 4% exhibit deep integration, with AI involved in 75% or more of their task portfolios. This indicates that, for most roles, AI is used selectively rather than comprehensively.
Augmentation vs. Automation: The study finds that 57% of interactions are augmentative — supporting human decision-making and iterative processes — while 43% are automation-focused, executing tasks with minimal human intervention. This split emphasises the dual role of AI as both a collaborative partner and an efficiency enhancer, providing a balanced picture of technology’s present and near-term influence on work.
Occupational and Economic Dimensions: Analysis by wage and educational requirements shows that AI usage peaks in mid-to-high wage occupations, particularly in roles demanding advanced computational and cognitive skills. Both lower-wage roles and very high-wage professions (which often involve specialised physical or clinical tasks) show lower levels of AI integration, highlighting a nuanced pattern of adoption across the economy.
These findings have significant implications for workforce development, suggesting that targeted upskilling and strategic policy interventions may be required to ensure that workers in varying sectors can benefit from — and adapt to — the evolving AI landscape.
Challenges and Limitations
Data Scope and Representation: The research is based solely on data from the Claude.ai platform, which may not fully capture AI usage across all industries or platforms. There is a possibility of selection bias in the types of tasks and occupations represented.
Static Occupational Classifications: The reliance on O*NET’s static descriptions means the study might not account for newly emerging tasks or roles that AI could create in the future or particularly unusual ways people use AI.
Interpretation of Task Usage: While the study identifies the presence of AI in task-related conversations, it does not provide insight into how AI outputs are used in practice, leaving open questions about the real-world effectiveness of these interactions.
Future Significance
Evolving Usage Patterns: As AI capabilities continue to develop, ongoing monitoring of task-level integration will be critical to understanding long-term impacts on productivity and workforce structure. Future studies could extend this analysis by incorporating data from additional platforms and by exploring sector-specific trends.
Policy and Training Implications: The findings underscore the importance of aligning educational programmes and training initiatives with the skills that are most augmented by AI. Policymakers might consider targeted interventions to support workers in sectors where AI integration is poised to transform job roles.
Anticipating New Tasks: The study provides a foundation for identifying leading indicators of how AI might create new economic tasks or redefine existing roles, suggesting the need for flexible regulatory frameworks that can adapt to rapid technological change.
Outcomes
This research is pivotal in offering one of the first large-scale, empirically grounded assessments of AI’s role in task performance across diverse occupations. By focusing on the granular use of AI in specific tasks, the study provides a clear framework for understanding and responding to the effects of AI on the economy.
Additional Reading
Brynjolfsson, E. et al. (2018). The Second Machine Age – Examining the digital transformation of the economy.
Frey, C.B. & Osborne, M.A. (2017). The Future of Employment: How Susceptible are Jobs to Computerisation? – A seminal study on job automation risks.
Acemoglu, D. (2021). The Work of the Future – An analysis of technological change and its impact on labour markets.
Eloundou, T. et al. (2023). Research on language models and their role in automating and augmenting workplace tasks.
Recent reports and surveys on AI adoption in various sectors, such as those conducted by Humlum & Vestergaard (2024) and Bick et al. (2024), which further detail AI’s real-world applications and economic impact.