The GenAI Divide: State of AI in Business 2025

The document "The GenAI Divide: State of AI in Business 2025", produced by MIT researchers under the NANDA project, presents the findings of a study on enterprise adoption of generative artificial intelligence. Based on the analysis of more than 300 public initiatives, 52 organizational interviews, and surveys of 153 executives, the report highlights a paradox: despite global investment exceeding $30 billion, 95% of companies fail to achieve measurable financial impact.

At the center of the analysis is the "GenAI Divide", a gap separating the few organizations that succeed in transforming processes with adaptive AI systems from the majority that remain stuck in pilots. The main cause is not model quality or regulation, but the absence of learning and memory in deployed systems. Widely used tools like ChatGPT or Copilot improve individual productivity but do not integrate into critical workflows. Attempts to create customized enterprise solutions often fail due to rigidity, lack of context, and poor alignment with operations.

The report identifies key patterns behind this divide: limited disruption in most sectors, large firms piloting but struggling to scale, investment bias toward visible functions like marketing instead of high-ROI back-office areas, and higher success rates for external partnerships over internal builds. It also highlights the “shadow AI economy”, where employees rely on personal AI tools to overcome official system limitations.

Successful projects share three features: deep integration into specific processes, continuous learning capability, and evaluation based on business outcomes rather than technical benchmarks. Companies that cross the GenAI Divide achieve clear benefits, especially in support automation, reduced external costs, and customer retention. The report also anticipates the rise of agentic systems and the future "Agentic Web", a network of autonomous agents able to coordinate processes beyond today’s limits.

In conclusion, the report is aimed at executives, innovation leaders, and professionals interested in understanding the real dynamics of generative AI in business. It provides a roadmap of risks and opportunities, emphasizes learning as the key barrier, and offers practical recommendations: prioritize buying over building, decentralize experimentation, and select solutions that adapt and evolve with the organization.

Key Points

• The report identifies the GenAI Divide, a gap between high adoption of generative AI and its low capacity to truly transform businesses.
• Despite enterprise investment exceeding $30 billion, 95% of companies see no measurable financial return from their AI projects.
• The main cause is not model quality or regulation, but the lack of memory, learning, and adaptation in deployed systems.
• General tools such as ChatGPT or Copilot improve individual productivity but fail to integrate into critical business workflows.
• Internal customized developments often fail due to rigidity, lack of context, and poor alignment with actual operations.
• Successful projects are defined by deep process integration, continuous learning capabilities, and evaluation based on business outcomes.
• A “shadow AI economy” has emerged, where employees use personal tools more effectively than official enterprise solutions.
• Companies crossing the GenAI Divide gain benefits in support automation, external cost reduction, and customer retention.
• Most sectors show little structural disruption, except for technology and media, which present clearer transformations.
• The report anticipates the rise of agentic systems and the Agentic Web, where autonomous, learning agents will reshape enterprise processes.