The engine

Alert engine & Smart Assist

Every stream watched by the same rules engine, every alert escalated to whoever is actually on shift, and an AI voice agent for the calls that would otherwise not get made.

Most control towers fail at the last step. The event is detected, an alert appears on a dashboard, and then nothing happens, because the person who needed to act was not looking at the dashboard at 2 a.m. Detection is the easy half.

The engine treats the alert as the beginning of a workflow rather than the end of one. It knows the alert type, the policy for that type, who is on shift, what to do when they do not respond, and when to stop trying. Where a call is the right action and no human is free to make it, the AI voice agent makes it.

Alerts & Smart Assist

The engine

Configurable alert types

Alert types are configuration, not code. Thresholds, severity, who is notified and what closes the alert are all set per customer, and a new alert type does not require a release.

Roster-based shift escalation

Escalation follows the roster round-robin, so an alert goes to whoever is genuinely on shift rather than to a static distribution list that was accurate a year ago.

Automatic closure rules

An alert whose condition has resolved closes itself under configured rules. Without this the dashboard fills with alerts nobody trusts, and an untrusted alert list is the same as no alert list.

Manual alert types

Breakdown, accident, no driver, preventive maintenance, driver on leave. Not everything worth escalating comes from a sensor, and the things a human reports belong in the same queue as the things a device reports.

Driving-pattern alerts

Night driving, continuous driving beyond a threshold, free-run and over-speeding — the patterns that predict an incident rather than report one.

OBD and ADAS monitoring

Harsh braking, acceleration, drowsiness, eye closure, seatbelt compliance and lane departure, arriving on the same queue and the same escalation path as everything else.

Smart Assist — the AI voice agent

Per-alert-type calling policy
Which alert types warrant a call, to whom, and in what order is configured per type and per customer. A drowsiness alert and a delayed-arrival alert do not deserve the same response.
Cooldowns and calling windows
The agent respects a cooldown between calls and a permitted calling window. An automated system that calls a driver six times in ten minutes destroys its own credibility.
Recordings and transcripts
Every call is logged with its recording and transcript, exportable. What the agent said and what the driver answered is auditable, which is the condition for trusting it with anything.
Scoped per customer
The voice agent runs per customer with its own policy set, so one tenant switching it on does not change behaviour for anyone else.
Alert dashboard showing volume by severity and a type-by-age matrix across night driving, long stoppage, geofence SLA breach, geofence skip detected and late departure

A live alert dashboard. The type-by-age matrix is the working view: what has been sitting unactioned, and for how long. Customer name masked; volumes and types are real.

Events the engine watches

Over-speeding — vehicle exceeds configured lane or zone speed
Harsh braking or harsh acceleration event
Driver drowsiness — ADAS eye-closure or head-drop detection
No seatbelt — ADAS seatbelt compliance failure
Geofence breach — unauthorised route deviation or restricted zone entry
Excessive idleness — vehicle stationary beyond configured threshold
Delayed vehicle arrival at pickup or loading point
Delayed delivery against committed SLA
SOS signal from driver or vehicle panic button
Temperature excursion in a reefer or cold-chain vehicle
Door-open event at an unexpected location
FASTag mismatch — expected toll plaza not transacted
Night driving and continuous driving beyond threshold
Skipped geofence — a checkpoint that should have been crossed
Where we draw the line

The voice agent handles structured, policy-bound calls — confirm, remind, ask a defined question, escalate. It is not a substitute for a control-room operator on an incident, and it is not positioned as one.

See it running on your operation

A working session over your own fleet, lanes, partners and SLAs — not a slide deck.