CAMBRIDGE, Mass. — Researchers at Massachusetts Institute of Technology have developed a new framework for analyzing where artificial intelligence can be applied across the economy, finding that AI use is heavily concentrated in a small share of work activities and is dominated by information-related tasks.
In a working paper published on arXiv, MIT researchers said they created a “comprehensive ontology” of work by reorganizing roughly 20,000 tasks from the U.S. Department of Labor’s O*NET database to better map how AI systems are used in real-world settings.
Using that framework, the researchers analyzed more than 13,000 AI software applications and an estimated 20.8 million robotic systems worldwide to determine how AI is distributed across different types of work.
The study found that AI adoption is highly uneven. According to MIT, the top 1.6% of work activities account for more than 60% of AI’s total market value, suggesting that AI is concentrated in a narrow band of tasks rather than broadly spread across all types of work.
72% of Focus
Most AI use is focused on information-based activities, which make up about 72% of total AI market value, the researchers said. Within that category, creating information accounts for roughly 36%, while only about 12% of AI is used in physical tasks.
Interactive work — which includes both digital and physical tasks — represents about 48% of AI usage, with a significant portion involving the transfer of information, according to MIT.
The findings suggest that AI is most applicable to “information work,” including tasks involving creating, analyzing and communicating data. However, MIT researchers said most occupations still contain at least some components that AI can assist with, reflecting the broad presence of information-processing tasks across jobs.
Differentiation Identified
The paper also distinguishes between two ways AI may reshape jobs: by automating tasks entirely or by augmenting workers as a collaborative tool. MIT said some occupations are more likely to delegate tasks to AI, while others will use AI to enhance existing workflows.
Overall, the researchers said their framework provides a systematic way to predict where current AI systems are most useful and how future advances could expand or shift those applications across the workforce.
