The challenge
Every fleet safety programme has the same ambition: catch the risk before it becomes an incident.
Most of them have the same problem too.
As fleet size grows, so does alert volume. Dashcams and telematics systems generate thousands of safety events every day, including harsh braking, speeding, distraction, and near misses. Each one needs to be reviewed, validated, and escalated if necessary. That work falls on a control tower team whose job is to watch, assess, and act.
For one large enterprise fleet operation, that team had grown to 20 full-time operators. They were processing thousands of alerts daily, every alert required manual review, and every review took minutes. The backlog kept growing.
The team was not failing. The model was.
How alerts were reviewed before
When a safety event fires, a control tower operator opens the footage, watches the clip, makes a judgment call about severity, logs the outcome, and decides whether to escalate. On a quiet day, that process is manageable. On a day when thousands of alerts arrive across a large fleet, it becomes impossible to prioritise.
Alert fatigue sets in fast. When everything is urgent, nothing is. Operators start making faster decisions, consistency drops, and high-risk events get buried under routine ones. The incidents that most need a human response become the hardest to find.
This operation was experiencing all of it. Despite 20 operators and a structured review process, response times for critical incidents were slow, accuracy varied between operators, and costs were significant. Adding headcount was not a solution that scaled.
The control tower needed to stop doing manual work and start doing meaningful work.
What Core9 deployed
Core9 deployed its AI computer vision and alert automation system across the control tower’s existing workflows.
The system did not replace the team. It changed what the team was asked to do.
Every incoming safety alert now passes through an automated filtering and classification layer before a human sees it. Computer vision validates the event in real time, confirming whether a genuine risk occurred, assessing its severity, and categorising it accordingly. Alerts that do not meet a risk threshold are resolved automatically. Alerts that require human judgment are flagged, prioritised by severity, and surfaced to the operator who needs to act.
What previously required 20 people reviewing everything now required 2 people reviewing what mattered.
The control tower did not become smaller. It became more capable.
The results
- 90% reduction in control tower staffing.From 20 operators to 2, without any reduction in safety coverage or response quality.
- Alert validation time reduced from minutes to seconds.Events that previously required manual clip review and judgment are now pre-validated by AI before reaching the team.
- Consistent, objective classification across the entire fleet.Human variability in alert assessment was replaced by a standardised AI model that applies the same criteria to every event.
- Faster response to the incidents that matter.With routine alerts handled automatically, operators are no longer buried. When a high-risk event fires, the right person sees it immediately.
- A control tower that scales without adding headcount.Fleet growth no longer requires staffing growth. The system absorbs additional alert volume without additional operational cost.
What the team does now
The 2 operators who now run the control tower are not doing less work. They are doing better work.
They are no longer watching footage of routine braking events that did not represent a genuine risk. They are focused on the events that require human judgment, human escalation, and human response. Their time is spent on prevention, coaching, and high-risk intervention rather than on manual review of an inbox that never empties.
The operational cost reduction is significant. The more important outcome, however, is structural. This organisation now has a fleet safety function that improves as alert volume grows, rather than struggling under the weight of it.
That is what the shift from manual monitoring to AI-led decision support actually looks like in practice.
Applying this to your control tower
Most fleet safety operations are running the same model this client was running before. They have a large team, a growing inbox, and a process built for a smaller fleet that has not kept pace with the operation it serves.
The gap between what your control tower is doing and what it could be doing is measurable. It is also closable.
Book a 15-minute call with the Core9 team to understand where the biggest gains are possible for your operation.