Understanding the AI economy

Summary

Google has launched the first version of its AI & Economy ATLAS study, an ongoing effort to map real-world use of its AI tools through 15 million aggregated and de-identified interactions drawn from the Gemini App, AI Mode, and the Gemini API. The dataset covers more than 150 countries, 140 languages, 800 occupations, and 4,000 tasks, providing an early empirical baseline for how people integrate conversational AI into both professional and personal activities.

The report shows that adoption spans nearly all sectors and reaches 68 percent of U.S. occupations that account for 90 percent of employment, yet remains selective within individual roles: on average, AI assists with only about 21 percent of tasks. Most interactions focus on collaborative support such as ideation, information retrieval, strategy, and learning rather than full task replacement, with fewer than 10 percent of workplace exchanges resulting in complete automation. Non-routine cognitive activities appear at roughly twice the rate seen in the broader economy.

Use extends beyond office environments. Workers in manual and technical trades, including automotive technicians and industrial mechanics, employ the tools for diagnostics, troubleshooting, and on-the-job learning, often favoring multimodal inputs. More than 86 percent of recorded interactions occur outside paid work, covering household administration and service navigation that standard economic statistics rarely capture.

Globally, per-capita usage closely tracks national GDP per capita, although several middle-income countries in South America and the Middle East show adoption rates comparable to wealthier economies. English accounts for only about one-third of conversations, indicating that users continue to work in their primary languages even for complex queries. The study relies on privacy-preserving aggregation and clustering methods developed by Google DeepMind to convert raw interactions into structured occupational and task categories while removing identifiable information.

Why it matters

This article provides empirical, large-scale data on actual AI adoption and usage patterns across the global economy. For Dutch researchers and policymakers, these insights are crucial for understanding workforce transformation, guiding AI integration strategies, and shaping evidence-based economic policies.