"80% of AI projects fail" traces to a footnote one line long (jamiewatters.work)

A recent investigation into the widely cited statistic that "80% of AI projects fail" reveals a troubling truth: the figure lacks robust empirical backing and can be traced back to a casual remark in a magazine article. While the RAND Corporation’s 2024 report acknowledges the failure rate in a hedged manner, it bases the claim on unnamed surveys of business leaders’ opinions rather than actual project performance data. This has led to rampant misattribution, with many conflating the figure with definitive research, thereby obfuscating the real challenges of measuring AI project outcomes. The significance of this unraveling is profound for the AI/ML community, emphasizing the importance of scrutinizing statistics before they propagate. In a field where accurate data drives investment and strategic decision-making, relying on unverified figures can misguide stakeholders. The article advocates for critical thinking regarding statistical claims, urging professionals to examine the metrics behind such figures—focusing on what is truly being measured and how. As the discourse on AI project success becomes increasingly mainstream, establishing a clearer, evidence-based understanding is essential to navigate the complexities of AI implementations moving forward.