What happened
Google launched a new interactive, open-access experience for its AI & Economy ATLAS dataset, allowing users to explore global AI adoption rates by occupation and country. Simultaneously, Google, Google DeepMind, and MIT FutureTech released new research analyzing how scientists use AI, based on 2,600 specialized models and a survey of over 600 scientists in the U.S. and U.K.
Google announced the launch of a new interactive, open-access experience for its AI & Economy ATLAS project. This platform allows users to explore millions of global data points regarding AI usage, including adoption rates across specific occupations like electricians and purchasing managers, as well as country-level adoption trends and home usage patterns.
Concurrently, Google, Google DeepMind, and MIT FutureTech published new research derived from ATLAS data. The study analyzed 2,600 specialized AI models and surveyed over 600 scientists in the U.S. and U.K. It utilized a new taxonomy from MIT FutureTech to map scientific tasks and AI usage patterns.
The research found that scientists use AI at higher rates than many other occupations, with nearly half using some form of AI daily. The study distinguished between the use of Large Language Models (LLMs) like Gemini, which are widely used across fields, and specialized AI models, which are more common in health and life sciences for domain-specific prediction and simulation tasks.
The study reported that scientists save just under seven hours per week using AI. However, it noted that this time savings does not immediately translate to new discoveries due to significant time spent validating AI outputs, an increased backlog of untested hypotheses, and bottlenecks in physical experimentation and clinical validation.
Why it matters
This release provides concrete, granular data on how AI is reshaping labor markets and scientific workflows, moving beyond generic adoption metrics to specific occupational and task-level insights. It highlights that while AI saves scientists nearly seven hours weekly, it creates new bottlenecks in validation and experimentation, suggesting that productivity gains require workflow redesign rather than just tool adoption. This offers a practical framework for organizations and policymakers to understand the real-world economic and operational impacts of AI integration.
The release provides a detailed, data-driven view of AI's impact on the global economy, specifically highlighting how different professions and regions are adopting the technology. This granularity is crucial for understanding the uneven distribution of AI benefits and challenges across the labor market.
The findings on scientific workflows offer a critical insight into the 'productivity paradox' of AI. While AI accelerates certain tasks, it introduces new friction points in validation and experimentation. This suggests that organizations must redesign their processes to fully realize AI's potential, rather than simply adding tools to existing workflows.
The collaboration between Google, Google DeepMind, and MIT FutureTech lends academic credibility to the findings, making it a valuable resource for researchers, policymakers, and industry leaders seeking to understand the practical implications of AI integration in high-stakes, knowledge-intensive fields.
What to watch next
Monitor how other industries adopt similar AI workflow redesigns to address validation bottlenecks. Watch for follow-up studies from the ATLAS project that expand beyond scientific fields to other specialized professions. Observe if the open-access data leads to independent academic analyses or policy recommendations regarding AI labor market impacts.
Future expansions of the ATLAS project to include more industries and geographic regions, providing a more comprehensive picture of global AI adoption.
Independent academic studies that utilize the newly released open-access ATLAS data to validate or challenge Google's findings on AI's economic impact.
Industry responses to the identified bottlenecks in scientific workflows, particularly in how organizations are restructuring their research processes to better integrate AI validation and experimentation.