AI Hype or Media Fail? The Real Story Behind the "95% Failure Rate"

Many headlines claim a new MIT study shows “95% of AI projects fail,” or that “the AI hype is over.” The number is attention-grabbing—but the story behind it is more nuanced.

Even the critics miss the bigger picture. Some focus on the 52 organizational interviews, but leave out that the research also looked at over 300 public AI initiatives and 153 survey responses from senior leaders. Not a massive dataset, but much more robust than they suggest.

What the report actually shows:

• About 95% of enterprise generative-AI pilots don’t deliver measurable impact on profit & loss.
• The real issues are poor integration, rigid workflows, tools that don’t adapt to feedback, and investments aimed at flashy use cases instead of higher-return opportunities.
• “Shadow AI” is widespread—employees adopting AI informally often generate more value than official top-down projects.

Sometimes I’m just amazed at how quickly people grab an eye-catching number and start writing—missing the real insights in the study.