Generative tools such as ChatGPT, Gemini, and Claude have made artificial intelligence (AI) highly visible in everyday work. Millions of people now use such tools to draft emails, summarize notes, translate text, or clean up presentations. This creates a powerful impression that AI has already entered workplaces at scale and that major labor-market effects are imminent. Yet the visibility of AI use and its transformative impact on work are not necessarily the same thing.
The forms of AI adoption most likely to transform productivity and employment are not necessarily the most visible ones: workers occasionally using a chatbot in a browser to save a few minutes on a document. Rather, they involve enterprises – across services, manufacturing, and agriculture – redesigning workflows, integrating AI into operations, and optimizing how work is allocated across teams. It is these organizational changes, rather than the occasional use of AI tools by individual workers, that are most likely to drive substantial productivity gains and labor-market effects. That kind of adoption is difficult and takes time; it requires organizational capability and high-quality data infrastructure, not just access to the technology or a subscription to a service.
