Understanding the AI Economy

Google has released AI & Economy ATLAS (Activity, Task, Landscape, and Adoption Study) v1.0. Based on 15 million anonymized AI interaction data points collected across Gemini App, AI Mode, and Gemini API, this is a large-scale study spanning 150 countries, 140 languages, 800 occupations, and 4,000 tasks.

Key findings are as follows:

• AI use in the workplace is broad but not deep: AI is used in 68% of occupations that account for 90% of total U.S. employment, but the actual tasks applied within a given job average only 21%.
• Collaboration-centered rather than automation: Most work-related AI interactions are for collaborative purposes such as brainstorming, strategy development, information search, and learning, with full automation accounting for less than 10%.
• Manual laborers also use AI: Workers in technical and manual occupations such as auto mechanics and industrial machinists use AI for diagnostics, troubleshooting, and on-the-job learning, and they use multimodal AI twice as much.
• Home use accounts for over 86%: More than 86% of AI interactions occur outside of work contexts, focusing on daily tasks such as administrative processing (taxes, licenses, civil complaints) and searching for appliance usage instructions.
• AI adoption rates correlate with GDP: Per-capita AI use by country is closely linked to income levels, but some middle-income countries in Latin America and the Middle East show adoption rates comparable to high-income countries.