Safety Technology Without Fake Confidence

Safety technology is useful when it changes risk in the real workplace. It becomes dangerous when it only changes the dashboard.

Singapore is actively encouraging technology-enabled Workplace Safety and Health (WSH). MOM’s WSH Technology page highlights examples such as electronic permit-to-work platforms, vehicular safety technologies, video analytics, drones and robots, health monitoring wearables, environmental sensors and heat stress solutions.

That is the opportunity. The risk is false confidence: management sees cameras, sensors, AI alerts or completed digital checklists, then assumes the work is controlled.

A dashboard is not a control if nobody acts on what it shows.

Why This Matters

Singapore’s WSH duties still sit with employers, occupiers, principals and other duty holders. Technology can support those duties, but it does not replace hazard identification, risk assessment, competent supervision, communication, maintenance and review.

This matters especially in higher-risk sectors such as construction, manufacturing, transport and storage, logistics, facilities management, marine and contractor-heavy work. These are exactly the places where dashboards, cameras, telematics, wearables and electronic checklists are attractive.

The practical question is not “Should we use technology?”

The better question is:

What risk is this technology supposed to control, and what happens when it detects a problem?

What Organisations Should Know

Safety technology should be assessed as part of the WSH risk management system. MOM’s risk management guidance reminds duty holders to identify hazards, evaluate risk, control and monitor risks, and communicate risks to stakeholders.

For any safety technology, organisations should be able to explain:

• What hazard or work activity it addresses.
• Where it sits in the hierarchy of controls.
• Whether higher-order controls were considered first.
• Who receives the alert, reading or dashboard signal.
• What action must be taken, and within what time.
• How the action is recorded and verified.
• What happens when the system fails, misses, over-alerts or is bypassed.
• What personal data is collected, who can access it, and how long it is retained.

This is especially important for workplace monitoring. MOM has stated that surveillance cameras and monitoring devices can support safety and security, but organisations must still comply with PDPA obligations where personal data is collected. Practical safeguards include clear rules, safety/security purpose, appropriate locations, worker notification, access control, storage and retention controls.

Common Gaps We See

Visibility Is Mistaken For Control

A camera may help detect a hazard. A sensor may show an unsafe reading. A dashboard may make risk visible.

But visibility is not the same as control. If nobody has authority to stop work, isolate equipment, evacuate, correct the condition or escalate the issue, the technology is only observing the risk.

Alerting Is Mistaken For Intervention

An alert that is too frequent, vague or unactionable becomes background noise. Supervisors may learn to ignore it, especially if they are already overloaded.

Every alert should have a defined owner, severity level, response time, escalation path and close-out evidence.

Digital Completion Is Mistaken For Field Verification

Electronic permits, QR inspections and digital checklists can improve discipline. They can also make weak controls look neat.

For high-risk work, digital records should be sampled against physical reality. Did the isolation exist? Was the barricade in place? Was the gas test current? Did the supervisor verify the condition at the workface?

AI Is Treated As Objective

AI analytics can help prioritise attention, but they depend on camera angle, training data, lighting, PPE, posture, workflow and thresholds. A score is not automatically truth.

For high-impact decisions, including stop-work, contractor control or discipline, AI should support accountable human judgement. It should not replace it.

Privacy Is Treated As A Side Issue

Worker monitoring can damage trust if it feels hidden, excessive or punitive. Safety purpose, access, use, retention and worker communication should be addressed before deployment, not after complaints appear.

Practical Steps To Consider

1. Start with the hazard, not the product.

   Avoid beginning with “we need AI cameras”. Begin with the real work problem, such as vehicle-pedestrian interaction at a loading bay or repeated hot-work permit gaps.

2. Apply the hierarchy of controls.

   Check whether the work can be eliminated, substituted or engineered to reduce exposure before relying on monitoring and administrative alerts.

3. Define the control logic.

   Document the trigger, recipient, authority, response time, evidence of action, verification method and failure fallback.

4. Pilot before scaling.

   Test the system in real site conditions, including lighting, dust, weather, vibration, connectivity, worker posture, PPE variation, false positives and false negatives.

5. Assign ownership.

   IT may own the platform, WSH may own the risk assessment, operations may own the work, HR may own worker communication, and supervisors may own response. If nobody owns the full control loop, the system will decay.

6. Measure control performance.

   Do not stop at counting cameras, alerts or completed checklists. Track response time, verified corrective action, repeat alert locations, system uptime, false-alert trends and evidence that exposure was reduced.

How DASH Consult Can Help

DASH Consult helps organisations evaluate safety technology through a practical WSH lens, not a vendor-led lens.

We can support:

• Technology-control mapping against risk assessments.
• Video Surveillance System (VSS) compliance and active-use review.
• Electronic permit-to-work and checklist verification.
• Worker monitoring and PDPA guardrail review.
• Sensor, wearable and dashboard implementation reviews.
• Alert response, supervision and contractor assurance workflows.

The goal is simple: safety technology should help control risk, not just make risk look organised.

FAQ

Does safety technology make a workplace compliant?

No. Technology may support compliance, but the organisation still needs proper risk assessment, control measures, supervision, communication, maintenance and review.

Is a camera the same as supervision?

No. A camera may support monitoring, deterrence and investigation. Supervision still requires competent people, clear responsibilities and timely action.

Should AI safety analytics be used for discipline?

Be careful. AI outputs should be understood, validated and reviewed in context. For high-impact decisions, human judgement and proper investigation remain essential.

What is the simplest test for safety technology?

Ask: “If this system detects a risk, who does what, by when, and how do we know the risk was reduced?”