Google’s AI and Economy ATLAS data shows India’s creative sector uses AI at 19% of work-related tasks, which is 1.6 times the global average. The United States leads in technical adoption, where computer and mathematical roles account for 30% of usage, double the rate seen elsewhere.
Google is releasing a new interactive tool to let users examine these figures across specific jobs, from electricians to purchasing managers. The platform tracks daily home usage and adoption rates by country.
Research from Google, Google DeepMind, and MIT FutureTech now details how scientists apply these tools. The study analysed 2,600 specialised AI models and surveyed more than 600 researchers in the US and UK using a new taxonomy from MIT FutureTech.
Which professions are using AI most, and where?
ATLAS data breaks down adoption by profession and region:
- AI by occupation: In OECD nations, computer and mathematical roles and business and financial operations lead usage. In non-OECD countries, office support, arts and media, and education roles rank highest.
- Income level vs. adoption: While adoption usually tracks with national income, Brazil and the UAE show higher rates than their GDP per capita would suggest.
- AI for manual tasks: Usage for real-time equipment diagnostics differs by region. Brazil and Germany see 7% of work AI usage applied here, compared to 4% in Japan.
How are scientists using AI?
Nearly half of the surveyed scientists use some form of AI every day. They employ both large language models and specialised tools for different tasks. Usage of models like Gemini is spread across many scientific fields, while specialised models are more common in health and life sciences, specifically for domain-specific data prediction and simulation.
Researchers report saving just under seven hours a week. This extra time does not automatically create new discoveries. Scientists spend considerable time validating AI outputs, which has created a backlog of hypotheses waiting to be tested. Bottlenecks are appearing in physical experimentation and clinical validation.
Like most jobs, science faces a gap between potential and output. Realising the full benefit of these tools requires redesigning scientific workflows to match the speed of new capabilities.
What it means
The data suggests that while AI saves time, it creates new verification work. The bottleneck moves from data gathering to physical proof. Teams must restructure their processes to handle the volume of new hypotheses generated.
