Although global enterprises have already invested more than $30–40 billion in the development of generative artificial intelligence, the vast majority of organizations are not seeing any tangible returns, according to the latest report from MIT Media Lab’s Project NANDA, titled The GenAI Divide: State of AI in Business 2025 (July 2025). 

Researchers point to a clear divide—referred to in the report as the “GenAI Divide”—between the massive scale of deployments and their actual impact on business operations. While 5% of pilot projects integrating AI are generating multimillion-dollar profits, as many as 95% show no financial effect at all. 

The report identifies the main barrier as the so-called “learning gap.” Most generative AI systems: do not retain prior information, cannot adapt to context, and fail to learn from incoming data. This results in fragile processes, weak integration with daily tasks, and a lack of lasting impact on organizational performance. 

Although more than 80% of companies have tested tools such as ChatGPT or Copilot, and nearly 40% report having implemented them, in practice these technologies primarily enhance individual employee productivity rather than contributing to company-wide profits. 

Meanwhile, “shadow AI”—the unofficial use of private AI tools by employees—has become widespread. The report reveals that while only 40% of firms have official AI subscriptions, as many as 90% of employees say they use AI at work on a daily basis. 

The authors emphasize that broad business transformation through AI will not occur until systems gain the ability to learn contextually and operate autonomously. Until then, the most visible impact of AI will be in reducing external costs rather than restructuring internal processes. 

The study drew on an analysis of over 300 AI projects, 52 interviews with representatives from various industries, and 153 surveys completed by executives at four international conferences. The full report is available in the NANDA Papers repository.