There is a growing consensus on the transformative potential of artificial intelligence (AI) for the global economy and the nature of work. Nevertheless, the resultant changes for employment, individual lives, and the economy as a whole are not predetermined or assured. Significant efforts are needed to guide AI's influence on our lives and workplaces in a positive manner. Effective collaboration among various stakeholders is essential for this endeavor, requiring a deep understanding of AI's integration and application within the economy. Society demands empirical insights and data-driven research to shape effective policies, programs, and measures.
To facilitate this understanding, Google has introduced the initial version of the AI & Economy ATLAS (Activity, Task, Landscape, and Adoption Study). This ongoing, large-scale study examines how users are engaging with Google’s AI tools in a de-identified manner. The first dataset (v1.0) aggregates 15 million human-AI interactions sourced from the Gemini App, AI Mode, and the Gemini API, all of which collectively serve over 1 billion users each month. The insights from ATLAS v1.0 encompass a diverse array of over 150 countries, 140 languages, 800 job roles, and 4,000 different tasks, making it the most extensive examination of real-world AI usage to date.
The findings from ATLAS reveal how individuals leverage Google’s AI technologies for various work-related and everyday tasks. The report offers an initial glimpse into a dynamic landscape: while AI capabilities are progressing, its adoption is also transforming, and mechanisms to measure its economic impact are still evolving.
What insights are emerging from ATLAS v1.0?
Here are several notable findings thus far:
Although AI is utilized across numerous industries, its application in workplaces remains somewhat superficial. It spans all sectors and covers 68% of job categories, which together account for about 90% of U.S. employment. However, AI is employed selectively within these roles; typically, it is utilized in only approximately 21% of job tasks.
This selective usage reflects a greater emphasis on collaboration and support rather than full task automation. ATLAS data indicates that most workplace AI interactions center around collaborative functions such as brainstorming, strategic planning, information gathering, and continuous learning. Non-routine cognitive tasks like creative design or hypothesis testing are engaged with AI interactions at a significantly higher rate—65% in work contexts versus 35% in the broader economy. Notably, fewer than 10% of these interactions result in complete task automation.
Importantly, AI is not confined to traditional white-collar positions; it is also aiding workers in manual and technical roles. For instance, occupations like auto technicians and industrial mechanics are utilizing conversational AI for real-time diagnostics and problem-solving. Remarkably, these workers are twice as likely to engage with multimodal AI, such as using tools to generate images or videos. An example of this application includes technicians leveraging AI to analyze intricate test outcomes, troubleshoot electrical issues, and assess machines for wear.
Beyond the workplace, AI is proving valuable at home, addressing high-friction administrative challenges that conventional economic metrics may overlook. An impressive 86% of interactions with AI tools tracked by ATLAS occur in non-work environments. People are discovering innovative ways to use AI for productive household tasks, such as researching products, managing appliances, and dealing with complex bureaucratic processes like taxes and licensing.
On a global scale, AI adoption is correlated with GDP per capita, albeit with striking exceptions. The ATLAS data reveals that AI is widely adopted across more than 150 countries and regions, covering 99% of the global population. The linguistic diversity in AI usage is also noteworthy, with English accounting for only about a third of interactions. Users are not abandoning their native languages even for complex tasks. A deeper analysis indicates that per-capita AI usage aligns closely with a nation’s wealth, raising concerns over a persistent digital divide. However, this trend is not absolute; some middle-income nations in South America and the Middle East are adopting AI at rates comparable to those in wealthier countries.
Additional findings from the study further elaborate on these trends, highlighting the continued evolution of AI in both workplaces and homes worldwide.
