Metamorphosis of Occupational Hygiene and the Hygienist

For decades, we have used a reactive approach based on empirical evidence gained through monitoring and risk assessments.  This approach, although valid and legally compliant, has inherent limitations which are widely acknowledged. The measurements capture snapshots from the time of sampling. Statistical analysis tools have given us additional insights into the data. However, neither comprehensively account for dynamic reality of workplace hazards across shifts, seasons and operational variations.

With the convergence of AI and IoT, we will evolve to be part engineer and part data scientist using big data analytics to fundamentally restructure the way we identify, assess and control occupational exposures. The industry will benefit from our goal of exposure prevention, instead of exposure management, through predictive exposure modelling and intelligent control systems.

We evolved from being early observers of workplace disease into today’s scientifically trained specialists who can anticipate, recognize, evaluate, control and confirm workplace hazards. Our growth reflected advances in science and technology. Therefore, we must continue to evolve and develop skills beyond conventional practices. These may include competence in Python/AI tools for data analysis, understanding of machine learning models, and the ability to communicate complex AI outcomes to non-technical stakeholders. The hygienist must become a data scientist validating algorithmic outputs against professional judgement.

The average Occupational Hygienist spends 35-50% of the time on administrative tasks such as data aggregation, documentation and reporting. In this flood of administrative effort, we are not just losing time, but “IMPACT”. Administrative clutter is robbing us of the time needed to step back and effectively assess risks, devise controls and mentor teams. By leveraging AI-driven automated agents, we can eliminate manual data entry from a scanned sampling record or a complex risk assessment and minimize report writing time. These tools help us instantly migrate the information or transform the data into any new format with 100% precision.

The integration of generative AI tools into such repetitive administrative tasks boosts productivity and most importantly frees up critical time for higher-value activities such as making occupational hygiene programs more effective, interpreting complex exposure patterns, designing novel control strategies etc. – jobs we are trained and employed to do.

Why AI cannot be an effective substitute

The Health and Safety at Work Act 1974 requires every employer to ensure the health, safety and welfare of all employees as work to so far as is reasonably practicable (SFAIRP). This is clearly insisting assessment about “proportionality”.

An AI agent may interpret that a certain exposure exceeds statistical limits, but it cannot consider the social, economic, and operational factors that define “reasonably practicable” control measures.

ISO 45001 and HSG65 emphasize risk assessment as a consultative process involving workers. AI can help us make informed decisions with superior data, but it cannot replace the social conversation that legitimizes risk decisions at workplaces.

AI in Control of Hazardous Substances

AI brings the most significant advancement extending beyond simple on/off control for engineering controls such as Local Exhaust Ventilation (LEV) systems. AI-enabled LEV systems can employ real-time sensor feedback from particulate monitors, pressure sensors, occupancy detectors, damper position sensors to modulate and balance the system dynamically. Machine learning algorithms can be used to analyse historical usage patterns, temperature data, production schedules to optimize airflow and achieve energy reductions. AI can be used to predict filter loading progression for scheduling maintenance optimally rather than relying on traditional time-based triggers. AI is also increasingly being used for enhanced simulation techniques by feeding operational data to optimize designs and authorize changes, without expensive on-site trials or prototypes.

A conventional annual monitoring at a paint spraying booth may indicate compliance with solvent Workplace Exposure Limits (WELs). But performance degradation of the paint booth due to filer loading, fan wear, dust accumulation in the duct etc. leaves unnoticed temporal gaps during which employee exposures can exceed the WELs. We can feed real-time data from sensors measuring VOC concentrations, Particulate matter, Pressure, Air velocity, Noise, Smoke patterns etc. into machine learning algorithms and identify exposure trends before they breach the WELs. Similar techniques can be used in pharmaceuticals, metal working, wood, automotive facilities to trigger maintenance before exposure increases.

AI-powered cameras connected to vision-based systems can be used in real-time to identify and mitigate unsafe workplace safety practices such as improperly worn or missing gloves, respirators, goggles, safety helmets, or awkward body postures during manual handling tasks.

Natural language processing (NLP) techniques can be used to analyse maintenance work orders to predict when work activities might cause specific hazards such as disturbance of Asbestos Containing Materials (ACMs) and then generate automatic permit-to-work triggers.

Smart wearables can be used to monitor heart rate variability indicating heat stress or physical overexertion, early warning for hypoxia in confined spaces, and Hand-arm Vibration (HAV) exposure triggers. These systems can be integrated with AI enabled analytics to anticipate adverse health effects before they manifest such as a construction worker’s physiological data generating an automatic heat stress alert 30 minutes before any signs show up.

Critical Concerns

Data Bias – AI models could have been trained on biased historical data. An AI model trained on exposure data mostly collected from male workers in a construction industry, cannot accurately predict musculoskeletal risk for female workers. We must therefore understand model limitations and recognize when AI predictions require verification through traditional methods to ensure accuracy.

Ethical oversight – The occupational health benefits of AI-connected wearables are undeniable. However, biometric monitoring can raise profound ethical concerns such as employees reasonably objecting to continuous physiological surveillance, perceiving it as invasive and degrading. The data generated when combined with location tracking and activity classification can be used for performance management or discriminatory employment decisions. Surveillance must be transparent and justified by genuine safety needs, not productivity improvement. The occupational hygienist’s role hence expands to include ethical oversight. This requires engaging workers in technology deployment decisions, establishing clear data governance frameworks, and ensuring monitoring enhances worker wellbeing rather than managerial control.

Transparent reasoning – Many advanced AI systems produce predictions without transparent reasoning. This opacity is problematic when AI makes safety-critical decisions. If an algorithm recommends downgrading PPE requirements based on predicted low exposure, can we justify this to workers without explaining the algorithmic logic? If an incident occurs, can we defend the decision to enforcement authorities or in litigation without understanding how the AI reached its conclusion?

The EU’s proposed AI Act classifies safety applications as “high-risk” requiring explainability and human oversight. I believe BOHS must adopt, liaison and lobby with the regulatory bodies to adopt similar principles – AI should augment, not automate, critical decisions.

Conclusion

The integration of AI into Occupational Hygiene practice is certain and, if supervised correctly, has transformative potential. Businesses will gain predictive exposure models, optimized engineering controls, and data-driven risk reduction programs. Workers will benefit from hazard identification forecasts and more effective protection. The profession will advance toward higher-value strategic roles, leveraging technology to extend our influence across more workplaces.

We must foster hybrid competencies covering established exposure science, data analytics, and expert understanding of AI tools. We must advocate “human-in-the-loop” approach, ensuring professional judgment continues to remain key factor to any critical decisions. We must engage with AI vendors and scrutinize transparency and bias in the models.

Occupational Hygiene will continue to remain a discipline rooted in context, judgement, and human behaviour. The occupational hygienist of the future will not be reactive but will actively predict and prevent risk. Our responsibility is to ensure they serve not only efficiency but justice and not only data but dignity. As Uncle Ben said, “With great Power comes great Responsibility”.