david cant

Will AI put health and safety professionals out of a job?

There’s a common stereotype when it comes to health and safety managers: the person in the hi-vis, carrying the clipboard, maybe a bit old-fashioned. You might see this person making their rounds, scrawling notes, ready to compile a big, dense file later on.

Don’t get me wrong, this is still the case in plenty of businesses across the UK and beyond, and for the most part, there’s nothing wrong with it. However, the experience should be highly valued, and a trained eye with a risk assessment can be truly powerful, assuming the risk assessment isn’t just filed away into a drawer after.

Every so often, I take a look around and think, ‘Wow, we’re really living in the future. Everything from phones to fridges is starting to look like props from Blade Runner.

And health and safety, too, is entering the future – whether you want it to or not.

The future of health and safety

Being able to stop accidents before they happen is the fundamental goal of risk management. The majority of your time as a health and safety manager is spent identifying the dangers and what can be done to prevent them. Unfortunately, the unpredictability of human nature (the ‘Human Factor’) means even the best-laid plans can go awry.

As I’ve discussed before, overcoming the Human Factor involves putting in the effort to know those working on your site like people, rather than just statistics. By tuning yourself in and identifying potential triggers for risky behaviour, you’ve got a better chance of tackling it.

But you’ve got a million and one thing to do. As much as you’d like to, daily briefings and chats with the team aren’t feasible (and, let’s be honest, they’ll get sick of it pretty quickly, no matter how fun you try to make them.) You’re also limited in the amount of data you can glean from even the most comprehensive risk assessments.

So then, having a magic calculation that can predict the future would be amazing, right? And that’s exactly what AI predictive learning aims to do: input some data and out pops all your answers. According to what you told the machine, there’s a 98% chance of a vehicle collision in the warehouse. A 74% chance inter-office politics could lead to a damaging increase in stress.

You sit back, relax, and watch your near misses and absences plummet while all this is happening. Sounds good.

But could it actually be bad news for you?

Predicting health and safety

Predictive analytics aren’t a new phenomenon in health and safety. For decades, health and safety specialists have tried different algorithms to predict risk management with varying degrees of success.

Even risk assessments are a type of predictive science: you’re inputting potential risk factors and identifying their level and severity. By doing so, you’re predicting the dangers before they appear and hopefully putting controls in place to prevent them. It might not feel like Minority Report, but it’s the same idea.

There are a few standard predictive models in the field of predictive analytics, which all offer variations on ways to forecast safety:

Classification

This is considered to be one of the most simple and widely used types of forecasting. Essentially, the algorithm you use classifies historical data that you’ve collected into various categories, allowing you to ascertain, for example, the likelihood of equipment failure if it’s not recertified or whether a particular department is more likely to suffer a work-related injury.

Forecast

This model takes historical data and assigns it a metric value, identifying the occurrence and regularity of past safety failures to predict the likelihood of future failures.

Outliers

Contrary to the two previous types, outliers modelling focuses not on existing patterns but anomalous data. By identifying anomalies and outliers, areas that need health and safety attention can be identified.

Limitations

Whilst all the above types of predictive modelling have their benefits; they also have their drawbacks. This can be overcome by using different aspects of each model and combining them – but the biggest drawback remains our little brains.

These models can only analyse small to medium data samples because, fundamentally, the human brain isn’t all that powerful. As a result, we struggle to see patterns and often fall into the trap of our own biases.

To allow larger sets of data to be analysed to provide real, in-depth predictions, you need a machine. You need artificial intelligence.

Machine learning in health and safety

Machine learning essentially describes the use of these algorithms without any human intervention. As a result, computers can analyse massive amounts of data far beyond human ability quickly and easily. This could be a serious boon for companies, which could massively improve health and safety predictions and minimise the impact of safety breaches with a little investment in artificial intelligence.

With the right data, an artificial safety assistant could identify the potential for machine failure before a single bolt comes loose based on past failures. Likewise, it could identify a clash between two personality types before office politics triggers a spiral of stress that impacts efficiency.

Most importantly, it could identify the smallest gaps in your safety control measures before it triggers a chain reaction leading to injury or worse.

It sounds like a dream for managers terrified of on-site injuries and their costs in human and financial terms. But as a health and safety manager, you might be wondering: where do I come into this, and is a robot about to take my job?

Not quite Blade Runner

Although artificial intelligence is undeniably powerful in predictive modelling, the good news is that we’re a long way from a dystopian future where health and safety professionals are a thing of the past.

Although humans remain one of the biggest drawbacks of truly effective predictive modelling, they also remain crucial. The effectiveness of these AI predictions relies on the quality of the data collected and inputted. Therefore, these models are most effective when health and safety managers work in tandem with them.

As a risk manager, you need to know your site and business. The machine – for now – relies on you to collect the right data from the right places. For all the talk of science fiction, even the most complex AIs are useless without a human on the other end telling it what to do and giving it the right information to work with.

AIs also need training, much like a human assistant. All businesses and industries are different, and what goes in one might not work for the other. A safety professional with a piece of real, personalised business knowledge is needed to ensure the machine models are accurate.

So, don’t worry about losing your job to the machines. Instead, take advantage of these impressive advances and think about how implementing machine learning into your risk management processes could benefit you. By gaining a deeper understanding of the science behind predictive modelling and streamlining your information-gathering, you can massively improve the safety of everyone under your care.

The potential for AI within health and safety is massive, but health and safety professionals remain a crucial part. So tell Harrison Ford he can stand down.

I’ve been working closely with businesses and health and safety managers for two decades to identify ways to improve their health and safety processes. Can I help you? Get in touch via the contact form below or message me on 07814 203 977.

contact david

Are you more of a phone person than a form person?

Send a message via text or WhatsApp 07814 203 977

David Cant David Cant portrait
  • This field is for validation purposes and should be left unchanged.