Is AI hyped?
Enterprises poured billions into AI. One report says 95% of pilots fail to deliver. What’s going wrong?
InsureStory
AI was supposed to be the golden ticket for business. By 2025, companies around the world had poured billions into it, especially generative AI. Yet a recent MIT report found that about 95% of enterprise AI pilots fail to deliver measurable value or revenue growth.
The scale of the problem is stark. Most pilots never get past experimentation. Companies lose millions on projects that don’t deliver. And while Microsoft, Amazon, Alphabet and Meta keep raising their AI spending, profits lag far behind.
Impressive in a demo, brittle in the real world.
The first reason is brittle workflows. Many models can’t adapt beyond the scenarios they were set up for, and they miss the cultural and operational nuance that drives real decisions. It’s like a satnav that only works in perfect weather: great in theory, useless when reality throws a curveball.
The second is misalignment. Too often AI is bolted on top of operations rather than built into the work people actually do every day, so it creates extra steps instead of removing them.
None of this means AI doesn’t work. It means the hype ran ahead of the hard part: clean data, clear problems, process change and people who trust the tools. For insurers, the lesson is to start small, pick problems with measurable outcomes, like claims triage or document reading, and scale only what proves its value.