Charles Gabriel already called it: 95% fail because they skip the problem
On a Brew Bytes episode recorded from NYC, Charles Gabriel put the number even higher, citing MIT research that '95% of all enterprise AI experiments are failing' and named the cause in plain terms. Real quote: 'they don't always know the problem that they're solving… they see cool AI tech and they want to bring it in and they want to pitch it,' followed by 'very poor deployment… not really aligning internally.'
What it actually means
Today's 88% headline is the same story told from the outside; Gabriel's is the operator's version from the inside. The failure starts before the model with a team buying a capability in search of a problem, then bolting it onto an organisation that isn't set up to run it. The fix isn't a cleverer agent. It's starting from the problem and the workflow, then choosing the tech to fit. If you're scoping an AI project this quarter, the honest first question isn't 'which model' it's 'what exactly breaks if this works, and are we ready to change it?'
Don't start with the AI. Start with the problem it's meant to kill.
Source: Brew Bytes — Ep 13, Charles Gabriel · Friday, 31 July 2026