Understanding the AI Economy (7 min read)

Google has released ATLAS v1.0 (Activity, Task, Landscape, and Adoption Study), a study analyzing how AI is being adopted into the economy and life, based on user data from over 150 countries worldwide. This research is based on 15 million anonymized interaction data points collected through the Gemini App, AI Mode, and Gemini API.

The key observations are as follows:

• AI Use in the Workplace: AI adoption is widespread across all industries and occupations (68% of jobs, representing 90% of total U.S. employment), but shows a 'wide but shallow' pattern, being selectively used in only about 21% of actual tasks.
• A Collaboration-Centered Tool: Currently, AI is primarily used for collaboration and assistance purposes—such as ideation, strategy development, information search, and learning—rather than for automating work. The proportion of task automation is under 10%.
• Expansion into New Occupations: AI is being utilized not only by white-collar workers but also as a tool for real-time diagnostics and troubleshooting for field technical workers, such as auto mechanics and industrial machinists, who are twice as likely to use multimodal AI.
• Value in Daily Life: Over 86% of interactions occur outside of work hours (in daily life), being used to support household activities or handle complex administrative tasks like taxes and licenses, creating value that existing economic indicators tend to miss.
• Global Adoption Status: While AI usage generally follows a trajectory similar to each country's GDP per capita, some middle-income countries in South America and the Middle East are adopting AI at a pace similar to high-income countries, demonstrating an exception to the digital divide.