The Challenge
The operator's NOC receives fault situations on its 4G/5G Packet Core (Nokia CMM and CMG) through BMC Helix. About 1,000 situations a quarter each cost 2 to 4 engineer-hours of manual root-cause analysis. Public models were ruled out on network-data confidentiality, and Nokia needed one design that Nokia, the operator and BMC would all sign.
What We Built
A multi-agent pipeline, the Situation Enrichment Agent, that turns each fault situation into ranked root causes and a recommended action. It is open-loop by design: the agent recommends, the NOC decides.
- A deterministic four-step LangGraph flow (Retrieve, Annotate, Recommend, Format) instead of an open agent loop, because the NOC needs reproducible, auditable output.
- Agent tools exposed over MCP, an A2A gateway for production, and an A2A integration with BMC HelixGPT's Best Action Plan.
- gpt-oss-120B with Gemma embeddings behind a LiteLLM gateway on the operator's Tanzu Kubernetes Grid, so all inference stays in the operator's tenancy.
- OpenTelemetry tracing end to end, and Entra ID OAuth2 at the gateway.
Outcomes
Six Packet Core use cases, three on CMM and three on CMG, in six months. Measured in the MVP across 250 historical situations: 87% alarm-noise suppression, a median root-cause analysis of 42 seconds against the 2 to 4 hour manual baseline, and 84% of annotations accepted as written by NOC engineers. The operator funded the move to production on these results.
The work is part of a TM Forum Catalyst on multi-agent intelligence for packet core operations, toward Autonomous Networks Level 4.