In August 2025, MIT issued the bold claim that 95% of AI pilots are failing.
In a report called The GenAI Divide: State of AI in Business 2025 by MIT’s Project NANDA, they observed that organizational learning gaps were causing AI investments to fall short, not the AI tech products themselves.
This correlates with data from 3Sixty Insights’ July 2026 benchmark report on service-enabled human capital management (HCM) technology, which found that only 13% of organizations fully use the platforms they’ve paid for, often due to inadequate implementation or upstream issues such as ineffective education and customer support.
There are a fair number of reasons to doubt this figure given what constitutes failure varies from organization to organization. However, other studies by IDC, S&P Global, and RAND report failure rates of approximately 88%, 42%, and 80%, respectively. Averaged out, roughly three out of four (76%) AI pilots are failing, stalling, or delivering little to no measurable positive impact.
When this is the environment AI initiatives are forced to operate in, it is no wonder they underperform. Like two ships passing in the night, both parties—the end-users and the AI vendors—fail to see eye to eye on how best to deploy and use it. So how could they possibly suddenly find value in stacking AI on top of that?
Continue Reading: