Driver-facing cameras have traditionally been used to record incidents and identify unsafe behaviour after it has happened. The latest generation of AI-based systems is aiming to move one step earlier: recognising patterns that may indicate an elevated risk before they develop into a collision. Motive has introduced three new capabilities—Lane Swerving Detection, Smoking Detection and Forward Parking Detection—designed to identify behaviours that can precede road incidents.

The lane-swerving function is particularly focused on fatigue-related risk. It identifies three or more swerves within five minutes when the vehicle is travelling above 50 mph and groups them into a single safety event. This gives fleet managers a clearer picture of repeated behaviour rather than treating every individual steering movement as a separate event. The system currently sends alerts to managers, with driver alerts planned for a future release.

The smoking detection function takes a different approach, looking for a cigarette in the driver's hand or mouth for more than five seconds while the vehicle is moving at at least 5 mph. It can issue an immediate in-cabin alert while also notifying safety managers. This illustrates how computer vision can identify relatively subtle behaviours that traditional vehicle data would not necessarily reveal.

Forward Parking Detection addresses another common but very different risk: low-speed collisions when a vehicle reverses out of a space after parking head-first. The system can notify managers about these events and provide drivers with information through the Motive Driver app, creating a feedback loop between detection and subsequent driver coaching.

What is particularly interesting here is the move from incident recording towards risk prediction. AI does not need to determine that a crash is about to happen with certainty to be useful. Recognising a sequence of behaviours associated with higher risk can provide an opportunity for intervention while there is still time to change the outcome.

This approach could become increasingly important as commercial fleets generate more video and vehicle data. The challenge will be ensuring that automated detection remains accurate enough to distinguish meaningful risk from normal driving variation, while giving managers useful information rather than simply generating more alerts.

The development points towards a broader evolution of AI in road safety: rather than only explaining what went wrong, intelligent systems may increasingly help identify when something is beginning to go wrong—and intervene before it becomes an incident.