A delivery van approaches traffic on Mombasa Road too quickly. The driver notices the slowing vehicles at the last moment, brakes hard and avoids hitting the car ahead by only a few metres. Nobody is injured, no vehicle is damaged and no insurance claim is made. From a traditional fleet-management perspective, nothing happened. From a road-safety perspective, however, the business may have just received one of its most valuable warnings.
That event is a near miss. It is the type of incident where an accident almost occurred but was avoided through braking, steering, another road user’s reaction or simple luck. Businesses that only investigate crashes can miss these warning signs completely. With modern AI dashcams and video telematics, fleet managers can begin examining the events that happen before an accident rather than waiting until damage, injuries or insurance claims force them to act.
This is becoming increasingly relevant as video telematics develops beyond ordinary recording. In October 2026, Teltonika introduced a new generation of AI-powered video telematics combining GPS tracking, livestreaming and on-device artificial intelligence capable of detecting road and driver events. The technology can recognise risks such as potential forward collisions, unsafe following distance, lane departure, driver distraction, fatigue and phone use, depending on the camera and configuration.
For a Kenyan fleet operator, the important change is not simply that the dashcam records video. A conventional camera may store hours of footage that nobody watches unless something goes wrong. An AI-enabled system can help identify particular events that deserve attention, allowing management to focus on a few important clips rather than searching through an entire working day’s recording.
Consider a truck that completes a Nairobi to Nakuru journey without an accident. On paper, it appears to have had a successful trip. The AI dashcam may nevertheless have recorded three forward-collision warnings, several occasions where the driver followed another vehicle too closely and one hard-braking event. Those events tell management that the journey may have been much riskier than the absence of a crash suggests.
This is why near-miss analysis can be more valuable than accident analysis alone. An accident shows that risk has already turned into loss. A near miss provides an opportunity to correct the behaviour while the vehicle, driver and customer relationship remain intact.
Fleet managers should still be careful with terminology. An AI dashcam does not necessarily display a message saying, “This was definitely a near miss.” Instead, the system may record individual safety events such as hard braking, collision warnings, unsafe following distance, lane departure or driver distraction. Management can review the event, video and surrounding telematics data and determine whether it represented a meaningful near miss.
A forward-collision warning is one useful example. Teltonika’s current Edge AI documentation describes a Forward Collision Warning function that uses the road-facing camera to identify potential collision risk ahead once the vehicle is travelling above the configured operating threshold. If a fleet repeatedly sees these warnings from one vehicle or driver, the pattern may indicate poor following distance, late braking or driving too quickly for the traffic conditions.
The important word is pattern. One warning does not automatically mean the driver is unsafe. Another vehicle may have suddenly changed lanes, traffic may have stopped unexpectedly or a pedestrian may have entered the road. Video provides the context needed to distinguish an unavoidable emergency from behaviour that requires coaching.
Headway monitoring can reveal another important risk. A driver may regularly follow the vehicle ahead too closely without ever causing a collision. From the driver’s perspective, everything may appear normal because this behaviour has become familiar. An AI dashcam can highlight repeated unsafe following events, allowing management to identify the habit before a sudden traffic stop exposes how little space the driver has left to react.
This is especially important on busy Kenyan highways where commercial vehicles may move between urban congestion and faster open-road conditions during the same trip. Following distance that appears manageable at low speed becomes much more dangerous at highway speed because the vehicle travels farther during the driver’s reaction time.
Heavy vehicles create an additional concern because their stopping characteristics differ from those of small cars. A loaded truck requires more careful speed and distance management than a lightly loaded passenger vehicle. A driver who repeatedly closes the gap too aggressively may be creating risk even when every journey currently ends without damage.
Hard braking events provide another useful near-miss indicator. A single hard brake may be completely justified. A child may run into the road, another motorist may cut across the vehicle or traffic may stop suddenly. The driver should not be criticised simply because the telematics system recorded strong braking.
Repeated hard braking, however, deserves investigation. If the same driver produces these events several times every week, management can review the associated video and ask whether the driver is following too closely, approaching junctions too quickly or reacting late to changing traffic conditions.
Modern dashcams can also combine camera analysis with motion sensors. Teltonika’s current AI dashcam documentation includes motion-based detection for hard braking, harsh acceleration, sharp turns, collisions and rollover events. Combining video with motion data gives fleet managers a much clearer record than either source alone.
A harsh-braking alert might show the time and location, while the video explains why it happened. Perhaps the driver was speeding toward stationary traffic. Perhaps another vehicle suddenly entered the lane. Perhaps the driver was looking at a phone and reacted late. The same numerical event can represent very different levels of risk depending on what the camera shows.
This is where driver monitoring systems become particularly useful. Modern driver-facing cameras can analyse behaviours such as eye closure, yawning, distraction, phone use and seat-belt status, depending on the hardware and configuration. Teltonika’s 2026 systems, for example, can generate driver-monitoring events for distraction, drowsiness indicators and phone use.
Driver distraction is especially important because the dangerous event may last only a few seconds. A driver looks down at a phone, the vehicle ahead brakes and the fleet vehicle narrowly avoids a collision. Without video, management may only see a hard-braking event. With driver-facing footage, the business may understand what led to the late reaction.
NHTSA continues to identify distraction as a major road-safety problem and said in October 2026 that driver-monitoring systems may help detect distracted drivers and alert them to keep their attention on the road. The agency is also researching how telematics can identify risky behaviours such as swerving and hard braking.
Fleet managers should not interpret this as proof that AI can perfectly understand every driver state. NHTSA itself is studying the reliability of driver-monitoring systems in distinguishing distraction, drowsiness and other driver conditions. AI alerts should therefore support human review rather than automatically determine guilt.
That distinction is essential for building trust with drivers. If employees believe every camera alert will automatically result in punishment, they may view the system as hostile surveillance. If management explains that video is used to understand safety events, improve training and distinguish genuine mistakes from unavoidable emergencies, the technology can support a much healthier safety culture.
A strong fleet programme should therefore investigate events, not assumptions. If a driver receives a distraction alert, management can review the clip. If the driver simply glanced toward a mirror or dashboard for a legitimate reason, the event may not require action. If the footage repeatedly shows phone use while the vehicle is moving, the pattern is much more significant.
Near misses can also expose route-related hazards. Suppose several different drivers generate harsh-braking events at the same junction. The problem may not be the drivers at all. The intersection may have poor visibility, confusing lane markings or frequent pedestrian movement.
Fleet managers can use GPS location to identify where events occur repeatedly. Over time, a map of near misses can show high-risk points along regular delivery routes.
This creates an opportunity to change operational behaviour. Drivers can be warned about the location, a different route may be selected or internal speed limits can be adjusted for that area. Instead of coaching one driver after one event, the business improves safety for the entire fleet.
NHTSA has recently highlighted the potential for telematics and connected-vehicle data to identify dangerous road sections and risky behaviour patterns. The same principle can be applied at individual company level, even if a fleet operates only ten or twenty vehicles.
Imagine a courier company whose vans make hundreds of trips through Nairobi every week. If the AI dashcam system repeatedly records harsh braking near one particular roundabout, management has useful information that may never appear in an accident report because no collision has yet occurred.
Lane-departure events can provide another warning sign. A vehicle may drift toward another lane because the driver is distracted, fatigued or simply correcting poorly through a bend. Teltonika’s current Forward ADAS functionality includes lane-departure warnings at relevant speeds.
One event may result from avoiding a pothole or obstacle. Repeated unexplained lane departures can indicate a more serious issue.
Video again provides context. A driver moving slightly across the lane to give space to a cyclist should not be treated the same as a driver repeatedly drifting because they are struggling to stay awake.
Fatigue-related behaviour deserves particular attention in long-distance fleets. Drivers may work early mornings, late nights or extended routes where tiredness becomes a safety concern. A driver-facing camera may detect indicators such as prolonged eye closure or yawning, depending on the system.
NHTSA’s 2026 work on drowsy driving notes that modern Driver Monitoring Systems can use eye tracking and other signals to identify possible fatigue, while also emphasising that detection is only the first step and that effective intervention remains important.
For a fleet manager, this means the alert itself is not the solution. The company still needs a policy. What should happen when repeated fatigue alerts appear? Should the driver stop at the next safe location? Should dispatch adjust the schedule? Should another driver take over?
Installing technology without a response procedure creates little value.
The same applies to phone-use alerts. If management receives repeated events but never follows up, drivers quickly learn that the alert has no practical consequence. A clear fleet policy should explain whether handheld phone use is prohibited while driving, how events are reviewed and what coaching or disciplinary process follows repeated violations.
The purpose should be prevention rather than simply punishment.
Near-miss analysis can make driver coaching much more specific. Instead of telling an employee, “You need to drive more carefully,” management can discuss a real event. The fleet manager can show that the vehicle was travelling behind another car, the following distance reduced, the driver looked away and a hard-braking event followed.
This type of coaching is much easier for the driver to understand because it refers to an actual situation rather than a generic safety lecture.
Drivers can also learn from positive examples. A near miss does not always reveal bad driving. Sometimes the video shows excellent defensive driving. Another vehicle cuts into the lane unexpectedly and the fleet driver reacts calmly, maintains control and avoids the collision.
Management should recognise this too.
If dashcams are used only to find mistakes, employees will naturally distrust them. If they are also used to identify strong defensive driving, verify driver accounts and protect employees from false accusations, the technology becomes more balanced.
This protective role can be extremely valuable after customer complaints. A member of the public may report that a company
