Is AI hiding our failure modes?
A pattern I keep noticing — the more capable the tool, the less we notice when it papers over the problem instead of solving it.
Every team I talk to has adopted at least one AI tool that makes the surface of their work look better. The question that doesn't get asked often enough: what did the friction it removed tell us?
The signal in the friction
Friction isn't always a bug. Sometimes it's the shape of an unsolved problem trying to surface. When a tool smooths it out before you can look at it, you lose the diagnostic.
This is why we still write down what breaks. Not to blame the tool — but to notice what it hid.
Before recommending changes, we spend time on the problem itself — because the shape of the question determines the shape of every downstream answer.
- Start with the failure mode, not the fix
- Let structure explain behavior before naming people
- Ship diagnostics as artifacts, not conversations
- Solutionize before the problem is written down
- Roll findings into a slide deck without the raw notes
- Treat one team's account as the whole story
"Ask better questions. The shape of the question determines the shape of every answer."
Something I keep noticing about AI and problem definition.
Every team I talk to has adopted at least one AI tool that makes the surface of their work look better. The question that doesn't get asked often enough: what did the friction it removed tell us?
Read the essay →