Tony Evans caught his AI getting a figure wrong mid-prep, on camera, in one sentence
Today's data says the fastest-growing AI failure is the silent one: the answer that looks fine and isn't. On Brew Bytes Ep 10, Are Companies Sleepwalking Into the AI Era?, Tony Evans described catching exactly that, in his own prep, that morning when he asked ChatGPT what percentage of the workforce is Gen Z now and in 2030, ending up with the same figure for both.
What it actually means
In his words: "using ChatGPT just this morning as I'm preparing some stuff and I know it's got a figure wrong. I'm asking what percentage of the workforce now is Gen Z, projecting forward to 2030, what percentage of the workforce is going to be Gen Z? It gives me the same figure for both and it's like, well, how can that possibly happen?… If I didn't know to read through this and think to myself, that's a pretty simple calculation… I would have given the wrong figure." The model just answered, and the only thing standing between a wrong number and a boardroom was one person doing arithmetic in his head. Tony's word for the companies not doing that was sleepwalking but handing people access to Gen AI and calling it a strategy. The practical version for anyone shipping AI: human intervention should be a habit and a test. Know what the right answer looks like before you ask and check the response without be complacent.
Nothing errored. It just answered and only a human doing the sums caught it.
Source: Brew Bytes — Ep 10, Tony Evans · Friday 14 August 2026