Mostly, conversations around connected workers begin with wearables like smart helmets, smart watch, panic buttons, and biometric bracelets. This is where most press coverage and vendor activity has happened. While the industry has been building the ‘connected worker’ concept around new hardware, it has overlooked a sensor that has already been mounted on nearly every wall of the jobsite: the CCTV camera.
The main purpose of installing CCTV cameras is security surveillance and not worker connectivity, so they do not generally get mentioned in conversations on connected workers. I see this as a category mistake, not a technology one. The technology is already present but what is missing is the layer of reasoning that turns a passive video feed into a live safety signal.
Context, Not Just Vitals
Wearables are effective at reporting bodily functions. A wristband, for instance, can flag a worker’s abnormal heart rate or an unexpected fall, but it cannot explain the reason behind it. A worker’s heart rate may spike from heat, exertion, or because a reversing excavator is closing in behind him. Either way, a biometric sensor alone cannot provide the exact cause of the change.
This is where vision AI (artificial intelligence) brings something fundamentally new and valuable. Camera-based monitoring doesn’t just detect bodies in motion. It can also provide actual context, like confirm PPE (personal protective equipment) compliance, or flag when someone enters a restricted area. This situational context is what no wearable can infer.
Both approaches are complementary, not contradictory to each other. The value is in how the two signals connect. When a wearable flag an abnormal heart rate in a confined space, that alert alone tells a supervisor little beyond a number. If the same signal triggers a check against the camera analytics already running on that zone, confirming occupancy time, proximity risks, and whether ventilation equipment is active, the alert stops being a raw biometric spike and becomes a specific one: a named location, an identifiable hazard, and a timestamp a supervisor can act on.
The wearable initiates the alert, the camera adds the context, and only the combined output reaches the supervisor. That sequencing, not the individual sensors, is what makes the system connected.
I believe the technology to make jobsites safer already exists in most cases. What is missing is the discipline to connect what is already there before buying something new.
A wearable device recognizes the specific functions of the body, while camera-based AI is useful when it comes to the external aspects of the work performed. Put together, both supplement a connected worker programme that captures both internal and external signal, instead of treating either one as the full picture.
However, all this doesn’t make vision AI an alternative to wearables and sensors. Dust, poor lighting, occlusion, camera placement, and the like are some real limitations of vision AI systems on a working site. These systems require deliberate configuration, training, and validation rather than assuming they will work without any extra efforts.
My point is not that the camera replaces the wearable. It is that most connected worker strategies start with a hardware purchase before they ask what the site already has. A camera already mounted on a wall, feeding into a system that can reason about what it sees, is not a new sensor to buy. It is an existing one that has been underused. Before adding another device to a worker’s body, it is worth asking what the infrastructure already on the wall could tell, if anyone had built the layer to listen to it.
About the Author: Gary Ng, CEO and cofounder of viAct, has spent more than 10 years advancing digital transformation in construction. Under his leadership, viAct’s AI‑driven computer‑vision technology is reshaping how jobsites are monitored—reducing accidents, preventing delays, and boosting productivity and margins. He also contributes to the next generation of industry talent as a visiting faculty professional at The Hong Kong Polytechnic University. HE can be reached at gary@viact.ai