MIT study: 95% of AI pilots yield no ROI, signaling a value shift

A new MIT study casting a sharp light on the progress and payoffs of enterprise AI is reverberating through markets and boardrooms. After years of frenetic investment, researchers found that roughly 95% of generative-AI pilots yield no measurable return, suggesting the industry is moving from an era of hype to a stricter, proof-driven phase. The findings come as major technology shares have pulled back from recent highs, testing whether the AI boom can deliver real profits or remains largely a speculative bet on future productivity gains.

The GenAI Divide State of AI in Business 2025, conducted by MIT’s NANDA initiative, analyzed more than 300 publicly disclosed AI deployments, surveyed 350 employees, and interviewed 150 business leaders. It concluded that only about 5% of programs achieved rapid revenue acceleration, with the vast majority delivering little to no impact on revenue or earnings. The study highlighted a recurring “learning gap” within organizations: pilots falter when companies fail to adapt workflows, processes, and cultures to AI tools. In many cases, the most meaningful ROI emerged not from front-line customer-facing applications, but from behind-the-scenes...