Rethinking health and safety in the age of AI

Chris Coote at Dexory explains how AI is starting to transform workplace safety, not just in warehouses, but across all high-risk environments

 

For all the regulations, training sessions and high-vis vests, workplace injuries remain stubbornly common. The UK’s Health and Safety Executive reported over half a million non-fatal injuries last year. At Amazon’s UK fulfilment centres alone, there were 119 serious incidents and more than 1,400 ambulance callouts between 2019 and 2024. These numbers tell a simple truth: the health and safety measures we’ve relied on for decades are no longer enough.

 

Clipboards, paper forms, and PPE have their place, but they were designed for a world of slower, smaller-scale operations. Today’s workplaces, which feature sprawling warehouses, fast-moving production lines, and vast logistics hubs, are too complex to manage with reactive methods alone. Checking compliance after an incident has occurred may satisfy a regulator, but it doesn’t prevent the injury. If we are serious about protecting people, we need to move health and safety from the reactive to the prescriptive and predictive. 

 

 

Why safety is stuck in the past

Walk through most industrial sites and the safety process feels frozen in the 1990s. Risk assessments are often annual exercises, inspections rely on people spotting visible damage under time pressure, and reporting systems depend on workers filling in forms, often long after the fact. In theory, these checks should create a culture of accountability, but in practice, they build a culture of compliance and doing the bare minimum to tick the box.

 

That culture is a big part of the problem, and the reason why modern health and safety needs to be revamped. If safety is framed as an obligation, it’s rarely treated as an opportunity to innovate. Technology has transformed nearly every other aspect of industrial operations, from robotics to supply chain planning, so why is safety being left behind?

 

 

The hidden dangers technology can see

With the AI systems of today, that balance is starting to shift. Unlike human inspectors, AI algorithms can process huge volumes of data continuously and without fatigue. Robots equipped with advanced sensors and computer vision systems can now scan facilities in real time, spotting risks that are often invisible to the naked eye.

 

Take racking in warehouses. Fractures in beams or dangerous loading patterns are easy to miss during a manual check, but can be picked up instantly by computer vision systems. Adjacent technologies can detect structural weaknesses in bridges, rail lines or heavy machinery before they reach a point of failure. These aren’t futuristic ideas. They are being deployed today.

 

The implications are significant. Instead of waiting for issues to compound, organisations can act on early warnings, scheduling targeted maintenance or adjusting processes before people get hurt. Insurance costs fall, downtime is reduced, and, most importantly, lives are protected.

 

 

Designing technology people will trust

Spotting risks is only half the battle when it comes to implementing AI into health and safety practices. The harder challenge is ensuring people trust and act on what the system is telling them. Anyone who has worked in industry knows the reality of the situation, and that staff sometimes ignore alarms or warnings because they’re inconvenient, overly frequent, or poorly explained.

 

That’s why designing AI for high-risk environments isn’t just a technical challenge; it’s a human one. The most effective systems are those built with operators in mind, providing insights that are timely, in plain language and give context. A message such as “This beam may fail within six months unless repaired” drives far more action than a cryptic code on a dashboard.

 

Equally, organisations must demonstrate that the technology has been validated and is reliable. When workers see that predictions consistently align with reality, for example, a flagged rack is confirmed damaged by engineers, confidence grows. Over time, AI becomes not an outsider telling people what to do, but a trusted partner in helping them do their jobs more safely and effectively.

 

 

What the future looks like

As AI matures, we can expect safety processes to become increasingly automated. Routine inspections, whether of racks, machinery, or site infrastructure, will be carried out by autonomous systems. Incident reports will no longer rely on people filling in forms but will be generated from continuous streams of sensor data. Safety managers will spend less time chasing paperwork and more time acting on live insights.

 

The growing fear is that this will eliminate the need for human roles in health and safety, but the reality is different. The judgment calls, ethical decisions and leadership required to foster a genuine culture of safety remain firmly human responsibilities. What it will change is the balance, as people will be freed from repetitive, error-prone tasks and empowered to focus on what matters most, which is protecting their teams.

 

 

A turning point

Health and safety has always been about more than compliance; it is about dignity and care for the people who keep industries running. The tools we’ve relied on for decades have taken us far, but the plateau in injury rates shows their limits.

 

By embracing technologies that can help us move to more prescriptive and predictive ways of working, businesses can make safety proactive rather than reactive, embedding protection into the daily fabric of operations instead of treating it as an afterthought. The organisations that lead on this front won’t just reduce incident rates, they’ll set new standards for what safe, responsible work should look like in the 21st century. 

 

---

 

Chris Coote is Director of Product at Dexory

 

Main image courtesy of iStockPhoto.com and Thanadon Naksanee