Charles Gabriel on the pod: why 95% of enterprise AI projects fail
On Brew Bytes Ep 13, Charles Gabriel cited MIT research that 95% of enterprise AI experiments are failing and pinned the cause not on the models but on the deployment. His diagnosis: teams 'don't always know the problem that they're solving,' and they skip the unglamorous data work. 'It's pipes. Pipes then people,' he said. (Quotes from the real Ep 13 transcript.)
What it actually means
This week's failures, a detector misfiring on writers, agents breaking things in production, half-built AI features are the retail version of Charles’ point. The tech is rarely what fails; the deployment is. Companies buy the shiny model before they've named the problem it solves, then wonder why it stalls. If you're standing up AI this quarter, name the problem first and get your data clean and accessible second. The model is the easy part.
The model is rarely what fails. The deployment is.
Source: Brew Bytes — Ep 13, Charles Gabriel · Monday 3 August 2026