A new study from the Massachusetts Institute of Technology’s NANDA initiative finds that 95% of corporate generative-AI pilots are failing to deliver measurable financial returns, leaving only 5% of projects generating rapid revenue gains. The findings are detailed in “The GenAI Divide: State of AI in Business 2025,” released on 19 Aug. Researchers analysed 300 public deployments, interviewed 150 senior executives and surveyed 350 employees. They estimate that US companies have collectively spent $35 billion to $40 billion on generative-AI efforts, most of which remain stuck in pilot stages. The report attributes the shortfall to poor integration with business workflows rather than model performance, and says back-office automation, not customer-facing chat tools, yields the highest returns. Approach mattered: projects that relied on specialised vendor solutions succeeded roughly two-thirds of the time, while internally built systems succeeded about one-third as often. Workforce effects are emerging mainly through attrition and the decision not to replace outsourced roles rather than large-scale layoffs. The bleak success rate reverberated on Wall Street; the Financial Times reported a sell-off in US technology shares on 19 Aug. as investors questioned whether the recent enthusiasm over artificial-intelligence-driven growth is excessive.