A MESA GROUP RESEARCH SUMMARY
Ninety-Five Percent of Pilots Return Nothing
What MIT found when it looked for the money in enterprise AI
MIT NANDA, The GenAI Divide: State of AI in Business 2025 · A Mesa Group summary
This is the most cited number in enterprise AI, and it is usually cited for the wrong reason. The percentage is not the finding. The cause is.
The finding
95% of pilots showed no P&L impact
MIT's NANDA initiative examined roughly three hundred enterprise generative AI deployments. Approximately 95 percent of them delivered no measurable impact on profit and loss.
The study was built on three separate sources: about three hundred deployments analyzed, roughly a hundred and fifty executive interviews, and about three hundred and fifty employee surveys. The finding held across all three.
What separated the five percent
The researchers named the split the GenAI Divide, and the line does not fall where most people assume. It is not a divide between companies with better models and companies with worse ones. Every organization in the study had access to the same commercial models.
What separated them was integration. Whether the tool actually entered the workflow. Whether the organization around it was ready to absorb what it produced. The report attributes failure to enterprise integration and organizational readiness rather than to model performance.
What that means in practice
The implication is uncomfortable for anyone hoping the next model release solves this. It will not. A more capable model dropped into an operation that cannot absorb it produces a more capable pilot that still returns nothing.
Ninety-five percent is not a verdict on the technology. It is a measurement of the distance between a tool that works and an organization that can use it.
The honest caveat
The study has drawn a methodological critique worth knowing. Critics argue it defined success narrowly, as profit-and-loss impact within a relatively short window, and therefore undercounted efficiency gains and pipeline effects that had not yet reached the income statement.
That critique does not overturn the finding. It sharpens how the number should be used. Ninety-five percent is the doorway into the question, not the whole answer. The answer is in what the researchers found on the other side of the divide, and that part is not seriously disputed. The barrier was organizational.
THE ORIGINAL STUDY
MIT Project NANDA. The GenAI Divide: State of AI in Business 2025.
This page is a Mesa Group summary of independent research published by the organization named above. Mesa Group has no affiliation with it. Read the full study at the link.
ABOUT MESA GROUP
Mesa Group is a professional services firm in the Command Mapping category. The firm maps decision authority across an organization's operations and builds the working systems that authority calls for, delivered with full client ownership. Mesa Point leads the mapping practice. Mesa Built leads the building practice.The firm's principle is constant: automation with authority.