Case study · AI · Automation
AI-Assisted Network Operations
Deterministic network automation combined with AI-assisted analysis and fault troubleshooting.
sentinel · incident-4821 · assisted troubleshooting
> collect: show interface counters (leaf-101..104)
> deterministic check: CRC errors > threshold on eth1/12
> ai analysis: pattern matches optic degradation (0.91)
> recommendation: replace SFP, verify peer — awaiting approval
ApproveRequest more data
Context
Network operations teams spending most of their time on repetitive validation and on reading verbose output during incidents.
01
The challenge
- Manual pre- and post-change checks were slow and inconsistent.
- Incident triage required experienced engineers to interpret large volumes of device and fabric output.
- Appetite for AI, but zero tolerance for AI making unsupervised network changes.
02
Our approach
- Deterministic automation for configuration, validation and data collection via Python and APIs.
- AI layer that reads structured operational data and produces evidence-backed triage and fault hypotheses.
- Advisory-only AI output; all changes pass through the same validation and human approval as manual changes.
03
The outcome
- Engineers spend their time on decisions rather than reading logs.
- Faster, more consistent fault analysis with the supporting evidence attached.
- Full accountability for change retained with the operations team.
More case studies