MIT study finds 95% of firms struggling to reap returns on AI investments
More than half of AI spending was channelled into these domains, but researchers found that human involvement remained critical, and automated systems alone failed to deliver.
ADVERTISEMENT
Despite the frenzy around artificial intelligence and its promise to revolutionise industries, a new study by the Massachusetts Institute of Technology (MIT) suggests that the vast majority of companies are failing to see any meaningful returns on their AI investments.
The report, titled “The GenAI Divide: State of AI in Business 2025”, found that nearly 95 per cent of firms which attempted to incorporate generative AI into their workflows have experienced no measurable impact on profits. This revelation comes even as technology giants such as Microsoft and Google continue to restructure workforces, laying off staff and increasingly turning to AI in pursuit of higher margins.
According to MIT’s findings, U.S.-based firms have collectively invested between $35 billion and $40 billion into generative AI initiatives. Yet, for most, the expected gains have not materialised. The research was based on 150 interviews with AI leaders, an examination of 300 AI applications, and a survey of 350 employees across multiple organisations. The study concluded that the majority of AI pilot programmes faltered due to “brittle workflows, lack of contextual learning, and misalignment with day-to-day operations.”
While AI tools demonstrated effectiveness in streamlining back-office functions, particularly administrative and repetitive tasks, their application in more strategic areas such as sales and marketing fared poorly. More than half of AI spending was channelled into these domains, but researchers found that human involvement remained critical, and automated systems alone failed to deliver.
In comments to Fortune, Aditya Challapally, lead author of the report, acknowledged that a minority of companies had achieved striking results, with some “seeing revenues jump from zero to $20 million in a year.” However, he emphasised that these were exceptions, noting a significant “learning gap” across most firms struggling to adopt the technology effectively.
The study further revealed that corporate leaders often blamed regulatory hurdles and the limitations of current AI models for the lack of results. However, MIT’s researchers pointed instead to flawed integration strategies within enterprises as the root cause. Consumer-facing AI products, such as chatbots similar to ChatGPT, were also observed to have little or no influence on overall business outcomes.
For the 5 per cent of firms that succeeded, the difference lay in adoption strategies and implementation practices. Organisations that purchased AI tools from specialised vendors reported a 67 per cent success rate, while those that attempted to build systems in-house achieved only 33 per cent success. MIT noted that customised solutions, coupled with proactive managerial involvement in driving adoption, proved far more effective in enhancing productivity and growth.
The report also pointed towards the future of enterprise AI, highlighting the experimentation with agentic AI systems designed to learn, remember and perform actions autonomously. These early trials offer a glimpse of how artificial intelligence could evolve within businesses, even as most organisations today continue to grapple with integration challenges and unmet expectations.
