A widely misunderstood statistic from a new MIT report suggests corporate artificial intelligence projects are failing, but the study reveals the opposite. Michael Nuñez reports for VentureBeat that a “shadow AI economy” is thriving as employees successfully use personal AI tools for their work, outpacing official corporate initiatives.
The study from MIT’s Project NANDA found that 90% of employees regularly use personal AI tools like ChatGPT for work tasks, although only 40% of their companies have official subscriptions. This grassroots adoption has been faster and more successful than the early corporate spread of email or smartphones.
The report clarifies that the frequently cited 95% failure rate applies specifically to expensive, custom-built enterprise AI solutions. These systems often fail because they are rigid and do not learn from user feedback. In contrast, employees favor consumer tools for their flexibility and immediate usefulness.
The research also highlights that the highest returns on investment come from automating back-office processes like customer service and document processing, rather than from high-profile sales and marketing applications. Furthermore, the report found that partnering with external AI vendors is twice as successful as building tools internally. The study concludes that AI is succeeding through individual employee adoption, even as corporate strategies falter.
