AI investments failing? 95 per cent of firms see no returns, says MIT

A new study from MIT has cast a sobering shadow over the AI gold rush, revealing that a stunning 95% of businesses are failing to see any meaningful return on their generative AI investments. The findings, published in a report titled “The GenAI Divide: State of AI in Business 2025,” suggest a significant disconnect between the hype surrounding AI and its real-world business impact.

The report, based on an analysis of enterprise AI initiatives and interviews with hundreds of executives and workers, paints a picture of widespread experimentation with limited tangible results. While companies have collectively invested billions of dollars in tools like ChatGPT and other large language models, their use has been largely confined to boosting individual productivity for tasks like drafting emails or writing code snippets. This hasn’t translated into measurable gains for a company’s bottom line.

The primary culprit, according to researchers, is a fundamental “learning gap” within organizations. Rather than carefully integrating AI into their core workflows to solve specific business problems, many firms are simply applying generic, off-the-shelf models in a superficial way. These models often lack the contextual learning and adaptability needed for complex, mission-critical operations. The study notes that even advanced AI models can only reliably complete about 30% of office tasks, requiring human intervention to handle the rest.

Interestingly, the report highlights that smaller startups are having more success with AI. These agile companies are focusing on a single, well-defined problem and using AI to solve it, often seeing rapid revenue growth as a result. In contrast, large corporations are spreading their investments too thinly, resulting in fragmented projects that fail to scale.

The findings come amid growing worker skepticism, with a recent survey showing that 62% of employees believe AI is significantly overhyped. The report suggests that until AI systems can retain feedback, adapt to context, and seamlessly integrate into daily operations, the technology will continue to fall short of the transformative promises made by its most enthusiastic boosters. The MIT study serves as a stark reminder that true AI transformation requires a strategic, focused approach rather than a simple rush to adoption.