Google’s ATLAS provides a comprehensive view of how people use AI at work and home.

Led by Scott Strand, AI & Economy Lead at Google’s Chief Economist’s Office, the study analyzes 15 million anonymized interactions from the Gemini App, AI Mode, and Gemini API, used by over 1 billion people monthly. The first dataset covers 150 countries, 140 languages, 800 occupations, and 4,000 tasks. Findings show AI is widely adopted—68 % of occupations and 90 % of U.S. employment use AI, but only about 21 % of tasks per job involve AI. Most workplace use is collaborative—ideation, strategy, information retrieval—rather than full automation. Even manual trades like auto technicians and industrial mechanics use conversational AI for diagnostics and troubleshooting, often employing multimodal AI.

At home, 86 % of AI interactions address household research, appliance use, and navigating government services. Global adoption tracks GDP per capita, yet middle‑income countries in South America and the Middle East match higher‑income peers. ATLAS, built with Google DeepMind’s OCT framework, protects privacy through rigorous de‑identification. The project will evolve with AI’s rapid growth, offering policymakers and businesses evidence to shape AI’s economic impact.