The useful context is spread across too many systems
Routers, switches, monitoring, tower notes, customer records, tickets, maps, configs, and operator memory all matter during an incident, but no single screen contains the whole operational picture.
Troubleshooting knowledge lives in people
Experienced operators recognize patterns quickly, while newer staff rebuild the same diagnosis from dashboards, CLI output, old notes, and customer reports.
Automation becomes dangerous at the exact moment it becomes useful
AI can prepare diagnosis and changes, but production networking needs exact target, expected state, rollback, approval, and post-change evidence rather than open-ended device authority.