How AI safety monitoring supports ISO 45001 compliance

ISO 45001 is the international standard for occupational health and safety management systems. If your organisation is certified (or working towards certification), you already know the core requirements: identify hazards, assess risks, implement controls, monitor performance, and demonstrate continual improvement.

What you might not know is how well computer vision AI maps to those requirements. Not as a replacement for your management system, but as a tool that makes several of the hardest clauses dramatically easier to satisfy with genuine, evidence-based rigour.

Here's a clause-by-clause look at where AI safety monitoring supports (and in some cases transforms) ISO 45001 compliance.

A quick primer on ISO 45001

ISO 45001:2018 replaced OHSAS 18001 and provides a framework based on the Plan-Do-Check-Act (PDCA) cycle. It's structured around ten clauses, with the auditable requirements sitting in Clauses 4 through 10. The standard is adopted as a national standard in over 70 countries, including New Zealand and Australia, and certification is growing rapidly: the 2024 ISO Survey showed a 37% increase in ISO 45001 certificates over the two-year period from 2022 to 2024.

The standard doesn't prescribe specific technologies. It sets outcome-based requirements and lets organisations choose how to meet them. That flexibility is exactly why AI fits so naturally: it doesn't conflict with any ISO 45001 requirement, and it strengthens several of the most critical ones.

Clause 6.1.2: hazard identification and risk assessment

This is the cornerstone of the standard. Clause 6.1.2 requires organisations to establish, implement, and maintain a process for hazard identification that is "ongoing and proactive." It must account for routine and non-routine activities, how work is organised, and past incidents.

The challenge has always been making hazard identification truly ongoing. In practice, most organisations conduct periodic risk assessments (quarterly, annually, or after an incident) and supplement them with manual observations during safety walks. Between those touchpoints, hazards go unidentified.

Computer vision AI changes this fundamentally. When cameras are continuously monitoring your highest-risk areas, hazard identification becomes genuinely ongoing. Forklift-pedestrian proximity breaches, exclusion zone violations, speeding events, and PPE non-compliance are identified automatically, 24 hours a day, across every shift. The system doesn't rely on someone being present to observe the hazard, and it doesn't depend on whether a worker reports it.

For an auditor reviewing your 6.1.2 compliance, the difference between "we conduct quarterly risk assessments and weekly safety walks" and "we have continuous, automated hazard detection across all high-risk zones, generating data on every safety event" is significant. The latter demonstrates a process that is genuinely ongoing and proactive, exactly as the clause requires.

Clause 6.1.2.2: assessment of OH&S risks

This sub-clause requires organisations to assess OH&S risks from identified hazards while taking into account the effectiveness of existing controls. It also requires assessment of risks related to the OH&S management system itself.

This is where data volume matters. If your hazard identification process captures 15 near misses per month through manual reporting, your risk assessment is based on a tiny, unrepresentative sample. You don't have enough data to know which zones are highest risk, which shifts have the most events, or whether your controls are actually working.

When inviol detects hundreds of safety events per week, you have a statistically meaningful dataset. Heatmaps show you exactly where risk concentrates. Trend analysis shows you whether a new control measure (a changed traffic route, an added barrier, a revised shift handover process) has actually reduced events. You can assess the effectiveness of existing controls with evidence, not estimation.

Clause 8.1: operational planning and control

Clause 8.1 requires organisations to plan, implement, control, and maintain the processes needed to meet OH&S management system requirements. This includes establishing criteria for processes, implementing control of processes in accordance with those criteria, and maintaining documented information.

AI safety monitoring supports operational control in two ways. First, it provides a consistent, documented mechanism for detecting when operational controls fail. If you've established an exclusion zone around a loading dock, the AI system detects and records every breach. You have documented evidence that the control exists, that it's being monitored, and that breaches are being addressed through coaching workflows.

Second, it supports Clause 8.1.3 (management of change). When you make a change to a process, layout, or work practice, the AI data shows you whether the change has improved or worsened safety performance. This is evidence-based change management, and it's exactly what ISO 45001 is looking for.

Clause 9.1: monitoring, measurement, analysis and performance evaluation

This is where AI delivers perhaps its strongest ISO 45001 benefit.

Clause 9.1 requires organisations to determine what needs to be monitored and measured, the methods for monitoring and analysis, when monitoring shall be performed, and when results shall be analysed and evaluated. It also requires the organisation to evaluate OH&S performance and determine the effectiveness of the OH&S management system.

The standard specifically requires both leading and lagging indicators. Lagging indicators (injury rates, lost-time days) tell you what already happened. Leading indicators (near-miss frequency, safety event trends, risk scores) tell you what's about to happen.

The problem most organisations face is generating leading indicators at scale. Manual near-miss reporting captures a fraction of actual events (as we've discussed in detail). Without meaningful leading indicator data, the "monitoring and measurement" requirement is met in letter but not in spirit.

With inviol, leading indicators are generated automatically. The system tracks safety events by type, zone, shift, severity, and time period. It produces the data that Clause 9.1 asks for: what you're monitoring (specific hazard types and high-risk zones), how you're monitoring it (continuous computer vision detection), when you're monitoring (always), and when you're evaluating (weekly coaching reviews, monthly trend analysis, quarterly management review).

For organisations preparing for audit, this is a strong position. You're not just meeting Clause 9.1; you're demonstrating a monitoring system that most auditors would consider best practice.

Clause 9.3: management review

ISO 45001 requires top management to review the OH&S management system at planned intervals to ensure its continuing suitability, adequacy, and effectiveness. The review must consider the extent to which OH&S objectives have been met, performance trends, risks and opportunities, and the adequacy of resources.

The reporting dashboards that computer vision AI generates are built for this purpose. A management review can draw on trend data showing risk reduction over time, comparisons across sites or shifts, the impact of specific interventions, and a clear picture of where risk still concentrates. This is a far more rigorous and data-driven management review than one based solely on injury counts and audit findings.

Clause 10: improvement

The final clause requires the organisation to determine opportunities for improvement and implement necessary actions to achieve the intended outcomes of its OH&S management system, including continual improvement.

Continual improvement is the heartbeat of ISO 45001, and it requires evidence. You need to demonstrate not just that you have a system, but that the system is getting better over time.

When you have 12 months of continuous safety event data, the evidence of improvement (or the identification of areas where improvement is needed) is clear, quantified, and auditable. Across inviol's customer base, organisations see an average 67% reduction in safety risk events and a 42% reduction in recorded incidents over three years. That's the kind of documented, measurable improvement that Clause 10 is designed to drive.

Clause 5.4: consultation and participation of workers

This clause requires organisations to establish processes for worker consultation and participation, including providing mechanisms and resources for these activities.

A coaching-first approach to AI safety data supports this directly. When supervisors use blurred, privacy-protected footage to run group coaching sessions during shift briefings, workers are actively participating in the safety conversation. They're reviewing real events, discussing root causes, and contributing to solutions. That's genuine participation, not a checkbox exercise.

How it maps to NZ and Australian obligations too

ISO 45001 doesn't operate in isolation. For organisations in New Zealand, the standard aligns closely with obligations under the Health and Safety at Work Act 2015, which requires PCBUs to proactively identify and manage risks. WorkSafe NZ guidance encourages the use of ISO 45001 as a framework for meeting HSWA duties.

In Australia, WHS legislation similarly requires proactive risk management, and ISO 45001 certification is widely used as evidence of compliance. In the U.S., ISO 45001 aligns with OSHA's recommended practices for safety and health programmes.

In all three markets, AI safety monitoring strengthens your position by providing continuous, documented, evidence-based hazard identification and risk management that goes beyond what manual systems can deliver.

Not a replacement, but a force multiplier

ISO 45001 is a management system standard. It requires leadership commitment, worker participation, documented procedures, and structured review. No technology replaces those elements.

What computer vision AI does is provide the data infrastructure that makes those elements more effective. Your hazard identification becomes genuinely continuous. Your risk assessments are based on evidence rather than estimation. Your monitoring produces meaningful leading indicators. Your management reviews draw on quantified trends. And your continual improvement is documented, measurable, and real.

If you're working toward ISO 45001 certification (or looking to strengthen an existing certification), book a demo with inviol and we'll show you how our platform maps to each clause. It won't replace your management system, but it will make your management system significantly more robust.

Frequently Asked Questions

Does ISO 45001 require AI or specific technology?

No. ISO 45001:2018 is outcome-based and does not prescribe specific technologies. It requires organisations to identify hazards, assess risks, implement controls, monitor performance, and demonstrate continual improvement. AI safety monitoring is one way to meet these requirements more effectively, particularly for hazard identification (Clause 6.1.2), monitoring and measurement (Clause 9.1), and continual improvement (Clause 10).

How does AI help with the "ongoing and proactive" hazard identification requirement?

Clause 6.1.2 requires hazard identification that is ongoing and proactive. Computer vision AI provides continuous, automated detection of safety events across all monitored zones and shifts, producing a far more complete picture of workplace hazards than periodic risk assessments or manual observations alone.

What ISO 45001 clauses does AI safety monitoring support most directly?

The strongest alignment is with Clause 6.1.2 (hazard identification and risk assessment), Clause 8.1 (operational planning and control), Clause 9.1 (monitoring, measurement, analysis and performance evaluation), Clause 9.3 (management review), and Clause 10 (continual improvement). AI also supports Clause 5.4 (worker consultation and participation) through coaching-based use of safety data.

Is ISO 45001 certification growing?

Yes. The 2024 ISO Survey showed a 37% increase in ISO 45001 certificates from 2022 to 2024, and the standard is now adopted as a national standard in over 70 countries. Industries including construction, manufacturing, logistics, and energy lead adoption.

Does ISO 45001 align with New Zealand and Australian safety legislation?

Yes. ISO 45001 aligns closely with New Zealand's HSWA 2015 and Australian WHS legislation. WorkSafe NZ and Safe Work Australia both recognise ISO 45001 as a framework for meeting legislative obligations around proactive hazard identification and risk management.