Summary
Google has launched an interactive, open-access experience for exploring AI adoption across occupations and countries through its AI & Economy ATLAS project. New ATLAS-based research reports that nearly half of surveyed scientists use AI daily and save just under seven hours a week, while validation and experimentation are creating new workflow bottlenecks.
Google has launched a new interactive, open-access experience for exploring data from its AI & Economy ATLAS project, alongside research on how scientists use artificial intelligence. The findings show that AI adoption varies substantially by occupation and region, while scientists report meaningful time savings alongside growing pressure on the later stages of research.
The ATLAS project brings together millions of global data points on AI use at work and at home. Its new visualisations allow users to examine adoption by occupation, country and task, including uses ranging from office work to equipment diagnostics.
AI adoption differs by occupation and region
In India, arts, design and media occupations account for 19% of work-related AI usage, which Google says is 1.6 times the global average. The pattern contrasts with the United States, where computer and mathematical occupations make up 30% of work-related AI usage—twice the share reported for the rest of the world.
The occupations with the highest usage also differ between groups of countries. In OECD countries, computer and mathematical work and business and financial operations lead AI usage. In non-OECD countries, office and administrative support, arts, design, entertainment, sports and media, and education and library occupations occupy the top positions.
Google reports that AI adoption generally rises with a country’s income level, although Brazil and the United Arab Emirates have higher adoption rates than their gross domestic product per capita would predict. Manual work is also represented in the dataset: real-time equipment diagnostics and troubleshooting account for 7% of work-related AI usage in Brazil and Germany, compared with 4% in Japan. The 7% figure is 1.4 times the global average, according to Google.
These comparisons show that AI adoption is not limited to software development or other highly technical jobs. The tasks being augmented differ according to local industries, occupations and economic conditions.
Scientists use both general and specialised AI models
The accompanying study from Google, Google DeepMind and MIT FutureTech analysed 2,600 specialised AI models and surveyed more than 600 scientists in the United States and the United Kingdom. The researchers organised the analysis using a taxonomy designed by MIT FutureTech to map the tasks scientists perform.
Nearly half of the surveyed scientists use some form of AI every day, according to the study. Scientists use both large language models (LLMs), such as Gemini, and specialised models, but the two types are used in different ways.
LLM use is distributed broadly across scientific fields and task categories. Specialised models are relatively more common in health and life sciences, as well as in domain-specific prediction, generation and simulation. This division reflects the different strengths of the systems: general-purpose models can assist across a wide range of language and reasoning tasks, while specialised models are designed around particular scientific data or workflows.
The surveyed scientists reported saving just under seven hours per week with AI. Google says that this time can be redirected towards research, but the productivity gain is not automatically translating into faster discoveries. Scientists also report spending significant time validating AI outputs. As AI makes it easier to generate hypotheses, more ideas accumulate than can immediately be tested, producing a backlog.
The study identifies further bottlenecks in physical experimentation and clinical validation. In other words, faster digital work can shift constraints to stages that still require laboratory experiments, real-world measurements or clinical processes.
Why the findings matter
ATLAS provides a way to compare AI use across occupations rather than treating adoption as a single global percentage. The India and U.S. examples also show how the distribution of AI use can be shaped by the kinds of work carried out in different economies.
For science, the findings point to a workflow problem as well as a productivity opportunity. AI can reduce time spent on some research tasks, but laboratories, experiments and validation may need to handle a larger flow of proposed ideas and model-generated results. Google describes ATLAS as a long-term research project and says future work will examine how AI is transforming the economy and how scientific processes may need to change.