An AI dash cam is a camera facing the road and often a second one facing the driver, with AI that spots phone use, tailgating or dozing, warns the driver at once, and keeps clips for coaching and evidence. Fleet cameras began as a way to collect evidence after a crash. The AI models on them now warn a driver in the moment and coach the driver afterwards.
The AI runs on the camera itself, which is called edge AI. This is needed because the warning must come within about two seconds, and a mobile connection on a highway cannot carry video fast enough for that. The camera analyses the video with video analytics, and sends only short clips of the risky moments to the cloud. Sending all the video would cost far more in mobile data than the fleet would pay for the camera.
Here is a typical case. A lorry driver on a highway near Nagpur picks up his phone. The driver-facing camera sees this, and a voice alert plays in the cab within two seconds. The camera saves a clip of about 20 seconds around the moment. It reaches the fleet manager, who watches it and later talks to the driver about it.
The camera differs from a telematics device in what it sees. The box reports that a driver braked hard. The camera shows why: a two-wheeler cut in front of him, so the braking was correct and the driver should be praised. The clip is also useful as evidence. After an accident, a fleet or its insurer can show what happened, which matters in claims in an Indian motor accident claims tribunal. Many Indian fleets buy the camera for this reason first.
The clips go to fleet management software, and the events are added to the driver's score in usage-based insurance, because fleets with fewer accidents can pay less for insurance. The idea of running software close to the device is explained in edge computing. The same cameras raise questions about watching drivers all day. Makers therefore try to show drivers that the camera protects them too, and some fleets blur the inside video unless an alert happens. The scene is fleet management.