When the Expert Retires: Cement’s Invisible Safety Risk
Walk any cement plant and the hazards announce themselves: a kiln burning hotter than molten lava, conveyors running for kilometers, confined silos, silica dust, coal dust, mobile equipment sharing lanes with people on foot. Those risks are real, and we treat them as real — we assess them, we write procedures, we lock them out. The risk that is not on any hazard list is the one I worry about most: the expertise that keeps people safe around all of it is retiring faster than anyone is capturing it.
A near-miss that ran on luck
I have watched a version of this play out more than once. A turnaround crew goes into a confined space — a silo, a duct, a tunnel — that a recently retired technician knew had a quirk: a spot where conditions were not quite what the standard entry plan assumed. He knew to check it. The crew that came in for the shutdown did not, because the quirk had never been written down anywhere. Most days, nothing happens. But “nothing happened” was riding on the hazard staying quiet, not on the system catching it. That gap — between what the experienced person knew and what the procedure captured — is the thing I worry about, and it widens every time one of them retires.
The procedure is not what keeps people safe
The written procedure matters, but it is not the thing that prevents the incident. The thing that prevents the incident is the operator who knows which task sequence has produced a near-miss before, which equipment configuration bites people when it is rushed, and which HAZOP finding was signed off but never really closed. That knowledge lives in heads, not binders. A pre-task risk assessment filled out to satisfy a requirement does not contain it. The person does.
The retirement wave is a safety event in slow motion
The operators, technicians, and EHS professionals who built their careers in cement through the 1980s and 90s are in the last stretch of their working lives. Their successors arrive with credentials but not context — and context is what safety runs on. When a senior kiln operator retires, the plant loses decades of knowledge about how that specific kiln behaves at the edge of its envelope. You cannot transfer that in a two-week handover, and most of it was never written down to begin with. It is a serious safety event that unfolds over years instead of seconds, which is exactly why it is so easy to ignore.
It gets worse where you can least afford it
Two pressures compound the gap. First, contractors — during a shutdown they are often the majority of people on site, arriving with varied training and no familiarity with site-specific hazards. Second, multi-site operators carry several national regulatory frameworks at once over a common ISO 45001 foundation. In both cases the safety knowledge that would manage the risk exists somewhere — in a person, in a past investigation, in a closed permit — but it is not retrievable at the moment it is needed.
And it is becoming a capital-markets question
There is a second reason this is climbing the agenda. Safety performance is no longer only an operational metric — it is a governance one. ESG rating agencies weight occupational health and safety as a primary factor for industrial companies, and ISO 45001 certification is increasingly expected by institutional investors, public-tender authorities, and the supply-chain requirements of large customers. An operator whose safety program quietly depends on a handful of soon-to-retire experts is carrying a risk that now shows up not just on the plant floor, but in access to capital, insurance, and contracts. The demographic problem and the ESG problem are turning out to be the same problem.
What AI can do here — and what it cannot
I want to be precise, because safety is the wrong place for hype. AI does not replace a safety professional's judgment, and it does not replace the physical walk-down where a human has to actually look at whether a machine is guarded and a space is safe to enter. What it can do is capture and structure the institutional knowledge — the hazard profiles, the past near-misses, the contractor history, the open findings — so that it is retrievable instead of resident in one person's memory. It supports hazard identification, job-hazard analysis, and training-gap review. It does not run the safety program. It keeps the program's knowledge from retiring.
The point
None of this is solved by hiring faster, because the people who carry the knowledge are the ones leaving. It is solved by capturing what they know while they are still on site and making it retrievable to whoever is on shift — which is exactly the gap I have spent my work trying to close. I laid out the full picture — the global scale, the competency gap, and where AI genuinely helps and where it does not — in a white paper, if you want the long version.
Decision support and education only — not a substitute for a qualified safety professional, physical inspection, or your site's safety management system.