The MIT announced that 95% of organizations saw no return on the US$30 billion spent globally on generative artificial intelligence. The figure is part of the MIT Project NANDA, which tracks corporate adoption of the technology.
The report points out that trying to solve the problem with more technical training did not work. Only 15% of learning and development leaders consider their organizations' leadership programs effective at teaching AI.
The problem, according to the SHRM, which covered the study, is that AI fluency has become a leadership issue, not an operational training one. When leadership does not understand what the technology can and cannot do, AI projects are poorly designed and poorly evaluated.
For those operating systems in production, the data confirms what is observed in practice: AI investment without management alignment generates projects that never leave the pilot phase. The money goes to infrastructure and models, but governance to measure results is lacking.
The paper believes the bottleneck is not technical. Operations and engineering teams can deliver working models, but if leadership cannot define success metrics or integrate AI into business processes, the return does not materialize. Training leaders, not just teams, is the next step.