How AI is Revolutionising Site Safety

Standing in a windswept Auckland carpark, looking down at what will become New Zealand's busiest train station, it's hard to imagine the invisible network of intelligence now watching over construction workers below. Behind the construction fencing, amidst the diggers, cranes, and fluorescent-clad workers spreading concrete, a quiet revolution is taking place.

Cameras equipped with artificial intelligence are learning to distinguish between a construction worker and a digger, between a signpost and a pedestrian, between safety and danger. They're becoming the digital guardians of New Zealand's most hazardous workplaces.

When Seconds Save Lives

For Murray Robertson, Chief Executive of Downer—the infrastructure giant employing 33,000 people across 300 sites in New Zealand and Australia—the stakes couldn't be higher. "We're constantly looking at ways we can use technology to advance how we deliver our work more efficiently and, of course, more safely," Robertson explains. "One of the key areas we focus on is the interface between our people and the heavy plant operating on sites."

The challenge is both simple and complex: how do you keep people and massive machinery from colliding when human supervisors can't be everywhere at once?

Enter Danu Abeysuriya, Chief Technology Officer and founder of RUSH, whose company has partnered with Downer to deploy R/VISION—an AI-powered safety monitoring system that's transforming construction site surveillance.

Beyond Traditional CCTV

"We deploy portable, self-powered camera units on site that are connected via satellite or 4G," Abeysuriya explains. "You point them in the general direction, and they continuously monitor what's happening. That information is transmitted in real-time back to our cloud system for immediate analysis."

The difference between traditional CCTV and AI-powered monitoring is stark. As Abeysuriya puts it: "We joke that CCTV allows you to watch someone steal your stuff from two weeks ago. An AI system allows you to act immediately and trigger a workflow."

This real-time capability is revolutionising how construction sites respond to safety threats. When the AI cameras detect issues—whether it's cars speeding past roadwork sites, workers without proper personal protective equipment, or dangerous proximity between personnel and heavy machinery—they instantly alert site managers via text message.

The Speed of Safety

The technology's impact becomes clear when considering response times. Traditional safety monitoring might take a day or longer to identify problems with traffic management plans. R/VISION reduces this risk window to just three hours, allowing site managers to quickly deploy additional signs, cones, or stop-go personnel.

"These systems reduce your risk window by presenting information in a really understandable way that you can act on immediately, or they can trigger workflows and begin acting on your behalf," Abeysuriya notes.

For Robertson, this instant communication fills critical gaps in site oversight: "The system will send a text message to our site supervisor. You've got to remember that some of these sites are quite long, so that person can't be sited everywhere. This essentially provides vision where he or she is not already based."

Teaching Machines to See

The sophistication of this technology lies in its ability to learn. While humans instinctively distinguish between a worker and a piece of equipment, between foreground and background objects, AI systems must be painstakingly taught these distinctions.

"A human can look at an image and automatically distinguish between a construction worker and a digger, between a signpost and someone walking their dog, and can make judgements about what's important and what isn't," the technology demonstrates. "By contrast, artificial intelligence needs to be taught all these things, often slowly and painstakingly."

However, once trained, AI can process vast amounts of data almost instantaneously, providing 24/7 monitoring that human observers simply cannot match.

Life-Saving Applications

The technology's most critical applications involve detecting life-threatening emergencies. The AI is being trained to recognise when someone—whether a worker or member of the public—suffers a medical emergency like a heart attack and collapses on site.

"Those are the kinds of things you want to know about straight away, and you want to be able to respond," Abeysuriya emphasises. "An AI system allows you to trigger immediate action within seconds and minutes."

For mature organisations like Downer, this capability integrates seamlessly with existing emergency response protocols, automatically triggering first responder workflows that can mean the difference between life and death.

Proven Results, Expanding Horizons

The R/VISION system has already demonstrated its value across diverse applications. The platform processes over five million licence plate recognitions monthly and has successfully transitioned from pilot programmes to full production deployments across traffic management, parking enforcement, and construction safety.

The collaboration between RUSH and Downer, supported by funding from Waka Kotahi's Innovation Fund, has validated the technology across multiple site types—from static locations with no traffic to highly dynamic roading sites.

The Human Element Remains

Despite the advanced technology, both Robertson and Abeysuriya emphasise that AI doesn't replace human judgement—it enhances it. "R/VISION has the potential to act as a crucial safety net, providing our teams with the right information to make informed decisions on-site," Robertson explains. "It doesn't replace the importance of human judgement, but rather complements it as another valuable tool in our toolbox."

A Vision for the Future

As construction sites across New Zealand become increasingly instrumented with intelligent cameras, the potential for preventing accidents, injuries, and fatalities continues to grow. The technology represents a shift from reactive incident reporting to proactive safety prevention—from watching disasters unfold to stopping them before they happen.

In those tunnels beneath Auckland, among the 700 workers completing the city's rail link project, cameras are learning to be guardian angels. And across New Zealand's construction sites, artificial intelligence is quietly revolutionising the most fundamental promise any employer can make: that every worker goes home safely at the end of the day.

The future of construction safety isn't just about better barriers or brighter vests—it's about intelligent systems that never blink, never tire, and never stop watching over the people who build our cities.

‍