US companies are losing 2.4% of revenue on failed AI projects

A study by AI governance firm Credo AI and Wakefield Research surveyed 300 US business leaders and finds organizations waste an average of 2.4% of revenue on failed AI initiatives — roughly $11.3M for a $500M company. The core problem: companies rush to deploy AI without proper governance (risk management, compliance checks, performance tracking), so models get built, fail to deliver, and get scrapped. Key stats: 60% of AI models are never deployed, 38% are retired within a year of deployment. Credo AI is a venture-backed startup (co-founded by ex-NSA engineer Navrina Singh) that sells tools to help companies govern AI — basically, they profit from the very waste they're measuring. The report self-servingly implies the fix is better governance (i.e., their product), but the headline number is a useful benchmark for the real cost of the current AI gold rush.