Artificial intelligence has become an unavoidable topic whenever the conversation turns to the future of the economy and work.
Everyone agrees the potential is massive, but understanding what is actually happening in practice, in people’s day-to-day lives, is a whole different story.
There has been no shortage of talk. What was missing was real data, at scale, capable of showing how AI adoption is actually playing out — not just on paper.
That is exactly the gap Google set out to fill with Google ATLAS 🚀
The project, which stands for Activity, Task, Landscape and Adoption Study, arrives as the most comprehensive study ever conducted on how real people are using AI at work and in everyday life. Worth noting right up front: this is a continuous, large-scale, and fully anonymized study designed to track a landscape that is constantly shifting.
And the numbers are impressive:
- 15 million aggregated and anonymized interactions analyzed
- Over 150 countries represented
- 140 different languages
- 800 occupations mapped
- 4,000 tasks cataloged
All of this drawn from anonymized data from the Gemini App, AI Mode, and the Gemini API, tools that together reach more than 1 billion users per month.
What ATLAS v1.0 revealed is going to surprise a lot of people, especially anyone who still thinks AI is only for office workers or white-collar professionals. 👇
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
AI adoption at work is widespread, but still shallow
One of the most interesting findings from Google ATLAS is the confirmation that AI has already reached virtually every corner of the job market. The study shows that adoption in professional settings spans all sectors of the economy and touches 68% of all occupations, which together account for 90% of total employment in the United States.
But there is an important nuance here. Despite that wide reach, actual usage within each role is still pretty selective. In a typical job, AI is used for only about 21% of tasks. In other words, people are still figuring out which moments the technology truly helps, and which ones where it does not make much of a difference yet.
This behavior reveals that we are at the beginning of a collective learning curve. AI is present in many places, but it is not deeply woven into work routines yet. It is an honest snapshot of a transitional moment.
Collaboration beats automation (for now)
Another finding that busts a pretty common myth: the vast majority of AI interactions at work are focused on collaboration and assistance, not on automating tasks. The ATLAS data shows that people are using AI primarily to brainstorm ideas, define strategies, look up information, and learn new things.
More creative and analytical tasks, such as design and hypothesis testing, which the study classifies as non-routine cognitive work, show up in work-related AI interactions at a much higher rate than they represent in the overall economy: 65% versus 35%. And here is the kicker — less than 10% of those interactions fully automate a task.
In practice, this means AI is functioning far more as a work partner than as a replacement. It helps people think, organize, and execute, but the person remains in control. This data point is critical for the debate around the future of employment, because it pushes back against the idea that full automation is already taking over the market.
AI is not just for people who work behind a desk
One of the most striking findings from Google ATLAS is the shattering of a widespread stereotype: the notion that artificial intelligence is an almost exclusive tool for knowledge workers, those who spend their days in front of a computer. The data paints a very different picture — and a far more diverse one than anyone imagined before a study of this global scale.
Workers in manual trades and technical fields show up among the active users of AI tools in the study. Automotive technicians and industrial mechanics, for example, are using conversational AI as a real-time partner for diagnostics, troubleshooting, and on-the-spot learning the moment a question comes up.
A fascinating data point reinforces this: when these professionals use Google tools, they are twice as likely to turn to multimodal AI — meaning they incorporate images and videos as part of the interaction. In practice, automotive technicians and industrial mechanics are using AI to interpret complex test results, solve electrical wiring issues, and inspect machinery for wear and tear.
This behavior has a direct impact on how researchers, businesses, and policymakers view the role of artificial intelligence in the economy. When the data shows that a mechanic is already interacting with AI in a functional and productive way, the conversation around professional development and workforce transformation needs to be rethought. ATLAS puts these groups at the center of the debate, no longer on the margins.
The biggest use of AI happens outside of work
If you think AI is mainly used in professional settings, brace yourself for a surprise. According to ATLAS, more than 86% of interactions with AI tools happen outside of work. Everyday life is where the action is.
People are using the technology in new and interesting ways that often do not even show up in traditional economic metrics. This includes productive household activities, like researching before making a purchase or getting help using appliances and tools around the house.
But the most valuable insight lies in what the study calls high-friction administrative tasks. You know those annoying bureaucratic chores nobody wants to deal with? Well, people are using AI to navigate government services — things like understanding taxes, dealing with permits, and sorting out fines. These are situations that cause a lot of headaches, and AI is delivering real value that standard economic indicators simply do not capture.
Global adoption follows wealth, but with exceptions
Google ATLAS is not just a user behavior study. In practice, it functions as a high-precision economic barometer. By mapping how artificial intelligence is being used across more than 150 countries and territories, representing 99% of the world population, the project offers an unprecedented view of how different economies are absorbing this technology.
The linguistic diversity confirms this global spread. English accounts for only about one-third of AI conversations worldwide, and users do not abandon their native languages even for complex tasks. This shows that AI is being genuinely adopted across very different cultural contexts.
When we look at the data more closely, though, a warning sign appears. On a per capita basis, AI usage closely tracks each country’s relative wealth level, which raises concerns about a persistent digital divide. Fortunately, it is not a universal rule. Some middle-income countries in South America and the Middle East are adopting AI at rates comparable to wealthier nations, suggesting the landscape may be more dynamic than it first appears.
How ATLAS was built (with privacy first)
A study of this scale only makes sense if it is done responsibly. The insights from ATLAS are generated by OCTO, which stands for Observation Clustering and Taxonomy Organisation, a Google DeepMind tool capable of transforming massive volumes of unstructured text — like conversations with language models — into organized and understandable information.
Privacy was taken seriously at every step. Beyond removing any personally identifiable information, Google added several extra layers of protection. The system automatically strips out potential references to sensitive data, removes all links between the anonymized data and original user records, summarizes the content of the texts, and groups those summaries into clusters that represent multiple people at the same time. Nothing points back to a specific individual.
Why ATLAS is different from everything that came before
Studies on the impact of artificial intelligence on the economy are nothing new. What changes with Google ATLAS is the scale and the source of the data. Instead of relying on self-reported surveys, where people say what they do, ATLAS starts from real, anonymized interactions from users who are already using AI right now, in the present, to solve concrete problems.
This methodological difference is massive. When someone responds to a survey about technology use, they tend to overestimate or underestimate certain behaviors. When the data comes directly from interactions with the tools, what you get is a faithful record of what is actually happening, without any perception filters. That makes ATLAS a highly reliable source for any serious analysis of AI adoption.
What comes next
It is important to understand that ATLAS v1.0 is just the beginning of a long-term project. AI capabilities keep expanding, people keep finding new uses, and research methodologies evolve alongside them. Many questions about AI and the economy still need answers, and that is exactly the work Google’s AI and Economy Research Program is conducting in partnership with academic researchers and other experts.
It is also worth noting that there is a much broader universe of economic AI use that does not yet appear in ATLAS. We are talking about products with billions of users, like Google Workspace, Google Translate, and AI Overviews, as well as enterprise platforms like Gemini for Google Cloud and Gemini Enterprise, and frontier capabilities in areas like agentic coding and world models.
The idea is for the project to evolve over time, incorporating new versions with updated data. This means that, for the first time, the research field around AI, the economy, and work will have a continuous, globally representative data source to draw from. The ultimate goal is clear: to offer a sharper understanding of this transformation for researchers, governments, businesses, and workers — helping everyone shape together the ways AI can support people in a positive way. And in the world of applied research, that is a pretty big deal. 🔍
