Artificial intelligence is no longer a distant promise or a tool reserved for major laboratories. It is being used in offices, creative studios, research centers, and manual tasks across different countries. New data from Google AI & Economy ATLAS shows that adoption varies by profession, region, and level of economic development.
AI is adopted differently in each country
The ATLAS project brings together millions of data points on how people use AI in their daily lives and at work. Google has just launched an open-access interactive experience to explore these trends by occupation, country, and type of activity.
The data reveals striking differences. In India, for example, occupations related to art, design, and media account for 19% of workplace AI use, a share 1.6 times higher than the global average. The creative industry is using these tools to generate ideas, produce content, and speed up design processes.
The situation is different in the United States. Computer and mathematical professions account for 30% of workplace AI use, twice the proportion recorded in the rest of the world. There, adoption is more closely linked to programming, data analysis, and other technical tasks.
Does this mean one country is more advanced than another? Not necessarily. AI is being integrated into the types of work that already have the strongest presence in each economy. In one place, it may boost programming; in another, it may support education, administration, or creative production.
Which professions are using artificial intelligence the most
In countries belonging to the Organisation for Economic Co-operation and Development, computer and mathematical occupations, along with business and financial operations, lead AI use.
In countries that are not members of the OECD, other areas stand out:
- Office and administrative support.
- Arts, design, entertainment, sports, and media.
- Education and library-related occupations.
The relationship between wealth and adoption is also visible. In general, higher-income countries use more AI. However, Brazil and the United Arab Emirates stand out as important exceptions, with adoption rates higher than would be expected based on their gross domestic product per capita.
The technology is not limited to digital tasks either. In Brazil and Germany, 7% of workplace AI use is related to equipment diagnostics and troubleshooting in manual activities. That figure is 1.4 times the global average. In Japan, by contrast, it accounts for 4%.
This detail helps challenge a common assumption: AI does not only work in front of a computer. It can also help identify faults, look up technical instructions, or assist people who operate machinery and equipment.
Scientists save time, but a new bottleneck is emerging
Google and Google DeepMind, in collaboration with MIT FutureTech, analyzed 2,600 specialized models and surveyed more than 600 scientists in the United States and United Kingdom. The goal was to understand how AI is used at different stages of research.
The main result is clear: scientists use AI more frequently than many other professionals, and nearly half say they use some form of AI every day.
Two types of tools are used in their work. Language models, such as Gemini, are used broadly across different fields and tasks. Specialized models have a relatively stronger presence in health and life sciences, especially for predictions, data generation, and field-specific simulations.
Scientists also report saving nearly seven hours per week. That time could be dedicated to developing questions, analyzing results, or designing new experiments. But saving time at one stage does not guarantee that the entire process will move forward at the same pace.
AI can generate hypotheses faster than laboratories can validate them.
The research found an increase in the time spent checking system responses and a growing backlog of hypotheses awaiting testing. Limits also emerge at stages such as physical experiments and clinical validation.
More productivity does not mean immediate discoveries
This phenomenon is not exclusive to science. When a tool speeds up one task, it can shift the pressure to another part of the process. A marketing team may produce more drafts, but then need more time to review them. A programmer may generate code quickly, but still has to test it and fix errors.
That is why AI's real impact does not depend only on how many people use it. It also requires redesigning workflows, training teams, and creating review systems that preserve quality.
A map for understanding the AI economy
The new ATLAS platform aims to offer a more concrete view of this transformation. Instead of asking only whether AI will replace jobs, it makes it possible to observe which tasks are changing, who is using these tools, and what differences exist between countries.
That perspective is more useful for workers, companies, and governments. It also helps prevent rushed conclusions. AI adoption does not happen the same way everywhere, and its value depends on the context in which it is applied.
The big question now is not whether AI will enter the workplace. It is already there. The challenge is learning how to incorporate it without confusing speed with progress—and without forgetting that people are still responsible for reviewing results, making decisions, and giving them meaning.
Original source
https://blog.google/innovation-and-ai/technology/ai/ai-economy-atlas-september-2026