Most corporate generative AI projects are failing, an MIT report finds

A new report from MIT’s NANDA initiative reveals that 95% of corporate generative AI pilot programs are failing to deliver significant revenue growth. Sheryl Estrada of Fortune reports on the study, “The GenAI Divide: State of AI in Business 2025”. The research identifies a “learning gap” and flawed integration into company workflows as the primary reasons for failure, not the quality of the AI models themselves. Lead author Aditya Challapally explained that generic tools often stall because they do not adapt to specific business processes. The study found that purchasing specialized AI tools from vendors is twice as successful as building them in-house. It also noted that the biggest return on investment comes from back-office automation, not sales and marketing, where most budgets are spent.