Intelligence · Phase 3
Anomaly Alerts
Live notification of unusual patterns: stock, calls, machines, customer queries. Instead of monthly reports — real-time intervention.
What it does
01
Time-series detection
Forecast-Engine check every series hourly/daily. Anomaly score per series.
02
Threshold-free
No rigid limits. The AI learns your normal — including seasonal peaks.
03
Escalation path
Anomaly → insight → recommendation → module proposal. Click from symptom to root cause.
How it runs
- 01
Hourly check
Stock, calls, utilisation — all in the watch loop.
- 02
Alert
At > 2σ deviation → notification to CEO (or Slack/email).
- 03
Explanation
LLM pulls context from the event log and writes a natural explanation.
What you get
- "Fertiliser A consumption in week 14 is 20 % over forecast — reorder?"
- Slack bot with a daily alert digest
- Escalation log for audit
Requires
- → Forecasting module + anomaly job.
In the example
< 5 min
AI → UI (synchronous)
> 2σ
default threshold