South African telco operator1,811 sites · 200,000 alarms/dayPowered by NexusAI

NOCSense — Network Operations Intelligence

Alarm feed × Topology × Correlation ML × NL assistant × Learning × AI = NOC intelligence

Projected ROI · Alarm noise elimination, auto-remediation, operations optimisation and continuous learning
90%Daily Opex reduction
~$1.2MMonthly Opex saved
{ How NOCSense works }

From raw signals to decisions

Data sources
Alarm feed200,000 raw alarms a day
Performance metricsReal-time network KPIs
Topology database85,000+ tables
Revenue context$2.85M daily across ~2,000 sites

NOCSense AI engine

Topology-aware correlation and remediation · TMF 642 / 842 compliant

StandardisationCorrelationRoot-cause MLNL assistantAuto-remediationContinuous learning
Insights delivered
Executive summaryNOC KPIs & noise cut
Site rankingsBy revenue impact
Root-cause analysisTrue failure patterns
Action plansResource + outcome
95.87%alarm noise eliminated at peak
70.2%incidents auto-resolved, no human touch
4.2 minfrom alarm to automated fix
Real-timerevenue-impact tracking per site
TMF 642standardised alarm compliance
{ The challenge }

What stood in the way

The situation before NexusAI — in the operator’s own numbers where they exist.

  1. 018,234/hr

    Alarm Storm

    Engineers drowning in noise — 8,234 raw alarms/hour at peak, with no prioritisation.

  2. 02—

    Root Causes Hidden

    Just 7 true failure patterns buried inside a 20 million undifferentiated alarm data set.

  3. 03—

    Legacy Data Silos

    Alert feed, performance metrics and an 85,000-table DB — all disconnected, no unified intelligence.

  4. 04—

    No Commercial Awareness

    Faults had no link to revenue impact — the NOC worked by queue order, not by business value.

  5. 05—

    Manual, Reactive Triage

    Every incident sifted by hand from the raw feed — analysts chasing symptoms, not the underlying fault.

  6. 06—

    Slow Resolution

    A multi-hour manual process from alarm to fix, with no learning loop to speed the next one.

{ The solution }

What NexusAI delivered

A live production pipeline on a South African telco operator's real alarm data — standardising 200,000 alarms a day to TMF 642, correlating them topology-aware and auto-resolving incidents. No systems replaced.

  1. 01

    Auto Data Standardisation

    AI agents map 200,000 alarms/day to TMF 642 — 84.1% core compliance, 100% required-field coverage, 98.7% type accuracy.

  2. 02

    Topology-Aware Correlation

    Temporal + graph + clustering ML group cascading failures — collapsing 8,234 raw alarms/hour into 115 actionable alerts.

  3. 03

    Root-Cause Detection

    Surfaces the true failure patterns behind the storm — 7 root causes isolated from a 20M-alarm data set.

  4. 04

    Natural-Language NOC Assistant

    Engineers ask plain-English questions; AI returns site rankings, root causes and action plans — in seconds.

  5. 05

    Automated Remediation

    Rules + learning engine auto-resolve 70.2% of incidents, with resource assignments and outcome projections.

  6. 06

    Continuous Learning

    Every human intervention feeds back as training data — auto-resolution improves +5.3% week-on-week.

{ Results delivered }

The outcome

95.87%Noise eliminated

8,234 → 115 actionable alerts per hour at peak load.

99%Root-cause accuracy

Root-cause analysis at 99% accuracy across a 20-million-alarm data set.

70.2%Auto-resolution rate

+5.3% week on week — the model learns from every engineer action.

4.2 minAverage resolution time

For automated incidents — versus a multi-hour manual process before.

100%TMF field coverage

84.1% core compliance and 98.7% data-type accuracy across the feed.

$2.85MDaily revenue monitored

Network revenue now tracked in real time, per site.

← Previous case studyRevSenseRevenue acceleration platform Next case study →Passive NetworkDetection & revenue recovery
Get started

Get results like NOCSense

Turn enterprise systems into an intelligent operating layer that understands your business and executes operations autonomously.

Request a demoTalk to sales
Live in production within 90 daysNo systems replacedCloud, on-premise or hybrid