AI deployment in management of land transport systems is no longer confined to back-office analytics. Increasingly, it is being used in vehicles and on roads to identify risks in real time, turning cameras from being passive video recorders into active safety and monitoring tools.
That shift is changing the strategic conversation for fleet operators. According to Richard Schubert, Group Chief Operating Officer, Cartrack Technologies: “AI is moving from the back office to the frontline — literally onto windscreens and into drivers’ cabins… Cameras are becoming sensors, video is becoming real-time intelligence, and systems that once documented problems can increasingly identify risks early enough for people to act.
One example is how Singapore’s Land Transport Authority plans to progressively roll out an AI-powered video analytics system using existing bus cameras to detect bus-lane encroachments, illegal parking, road defects and non-compliant road works. Pilot testing has so far been achieving an average accuracy of 90%.
Elsewhere in South-east Asia, public authorities and corporations are experimenting with similar ideas in different forms.
- In Hanoi, authorities said an AI camera system covering key streets and intersections had, in its first month, recorded and verified 6,351 cases eligible for violation notices, while also supporting traffic signal control and congestion management.
- Thailand, the Philippines, Indonesia were participants in pilot projects where dual-facing AI dashcams provided in-cab voice prompts when risky behavior was detected; involving a cloud-based video-intelligence platform that surfaced alerts and patterns for targeted coaching and safety management.
Together, these examples point to a wider regional shift: AI is moving from retrospective reporting towards operational intervention. For fleet operators, that raises a more strategic question than whether the tools are technically impressive. It is whether organizations can use them to reduce risk without creating new problems around privacy, trust and over-monitoring.
Turning traffic management in preemptive mode
Schubert argues that fleet safety is one of the clearest examples of this change. “A traditional dashcam answers one question: what happened?” he said. “The more interesting question for the next generation of road-safety technology is different: can a risk be identified early enough for an intervention before it becomes an accident?”
That is the sales pitch from many telematics and video-analytics vendors, but the underlying operational logic is not unreasonable. Driver-facing and road-facing systems can now be used to flag driver patterns such as distraction, phone use, fatigue, lane departures and unsafe following distances in real time, while building a longer-term record of recurring patterns. For transport managers, the attraction is not just footage after a crash, but earlier warning and more structured coaching.
Still, the effectiveness of these systems depends less on the camera alone than on the surrounding process. Real-time alerts may help some drivers self-correct, but poorly designed deployments can create alert fatigue or damage morale. Data can support coaching, but it can also become a blunt instrument for discipline if governance is weak. That makes implementation a management issue as much as a technology purchase.
Preserving balance in AI surveillance
Schubert acknowledges part of that tension. “The purpose of safety technology should not be to watch drivers for the sake of watching them,” he said. “Its value lies in its ability to act as a digital co-pilot and provide an additional layer of awareness when human attention inevitably fluctuates.”
That framing will appeal to operators trying to avoid a “surveillance first” culture, but it needs testing in practice. Questions about data retention, access rights, false positives, driver consent and how alerts feed into performance management remain central. The same system can be experienced as a safety aid or as workplace monitoring, depending on how policies are written and enforced.
The business case is also broader than fuel savings or maintenance. “The question should not only be, ‘How much will this technology save us?’ but also, ‘What could it help us prevent?’” Schubert opined. That is a more useful starting point than simple returns-on-investment arithmetic. A serious incident can trigger repair costs, operational disruption, insurance escalation, legal exposure and reputational damage, especially for smaller fleets with little spare capacity.
Three strategic implications
For operators assessing AI-enabled safety systems, three strategic questions matter.
- First, what problem is being solved: crash reduction, liability disputes, compliance, driver coaching or all four?
- Second, how will video and behavioral data be governed, and who gets access?
- Third, what organizational changes will sit around the technology, from rest policies to training and manager follow-up?
Transport system management is becoming one of the clearest tests of whether AI can deliver practical value in the physical world. The harder question is not whether cameras can now see more, but whether organizations can use that visibility responsibly enough to improve safety without eroding trust.
