AI is moving into the operational core of health and safety. Cameras detect people near mobile plant. Systems flag unsafe behaviour in real time. Some tools are beginning to stop equipment, change how it operates or direct people away from a hazard. These tools are potential game changers for safety, and they also bring real challenges.
At the October Executive HSE Forum, hosted by Laing O'Rourke, senior HSE leaders discussed what this shift means in practice. One point came through clearly. As AI-enabled tools become part of how work is controlled, they become a governance matter for boards and executives. Safety professionals, executives and boards need to weigh both the benefits and the challenges before adopting them.
From monitoring to decision-making
Most safety technology to date has observed and reported. A person reviewed the data and decided what to do.
The next generation of tools will act. When a system decides to stop a machine or redirect a worker, the rules it follows become part of the organisation's safety management system. Someone has written those rules. Someone has decided what counts as "too close", when an alert becomes an intervention, and what happens when the system fails.
Boards need to know who those people are, and how their decisions are approved and reviewed.
Five governance questions for boards
1. Who owns the rules?
The logic inside an AI-enabled safety tool must align with the organisation's own procedures, risk assessments and operating rules. Forum members noted that rules coded into a system can drift from how work is actually done. Boards should ask who sets the rules, who approves changes and how they are kept consistent with other business systems.
2. How do we know it works?
A system that produces constant false alerts trains people to ignore it. A system that misses real events creates false confidence. Members described operators bypassing devices when alarms became a nuisance. Boards should expect evidence of performance in the organisation's own environment, beyond the vendor's claims.
3. Is it a control, or assurance?
Forum members were consistent on this point. Technology should sit within sound design, clear operating rules and effective controls. Organisations that treat a detection system as the primary control risk weakening the layers around it. Boards should understand where each tool sits in the hierarchy of controls.
4. How is worker data handled?
AI-enabled tools often rely on cameras and location data. This raises questions of privacy, workplace surveillance obligations and trust. Some systems process images on the device and store no video, which reduces the risk. Boards should know what data is collected, where it goes, who can access it and how workers have been consulted.
5. Does it reach the people most at risk?
On many sites, the plant and the people near it belong to subcontractors or tenants. A principal contractor or site owner may have limited ability to require a particular system. Boards should ask how technology investment extends across the supply chain, and whether procurement and contracting models support it.
Officers' due diligence
Under Australian WHS laws, officers must exercise due diligence to ensure the organisation complies with its health and safety duties. That includes understanding the hazards and risks of operations, and making sure appropriate resources and processes are in place.
As AI-enabled tools take on a role in controlling risk, they fall within that scope. Directors need confidence that the tools are governed, tested, reviewed and aligned with the rest of the safety management system.
Where to start
Map which AI-enabled tools are already operating in your organisation, including those used by contractors.
Identify who owns the rules and decisions within each tool.
Ask for performance evidence from your own sites.
Include AI-enabled safety tools in your board's regular safety reporting.
Review how your procurement model supports or discourages safety technology.

