I have reviewed safety programs that were, on paper, excellent. Written to the standard. Signed, dated, reviewed annually. And then I walked the floor and found a guard zip-tied open, a lockout procedure for a machine that had been replaced two years earlier, and an operator who had never heard of the program that governed his job.
None of that is visible from a desk. It is not visible in a document review, it is not visible in a metric, and it is not visible on a call. It is visible when you walk out there and look.
The floor is the only source of truth
Every safety document is a claim about reality. The program says energy is isolated this way. The training record says this operator was trained. The inspection log says the area was checked on the fifteenth. Each of those is a claim, and the only way to test a claim is to go see.
I am the first person to tell you that AI can do enormous amounts of safety work — document review, cross-referencing, program drafting, log reconciliation. My whole practice is built on that. But there is a hard line, and this is it. A model can tell you your lockout program is missing three required elements. It cannot tell you that the third-shift maintenance tech has a personal set of keys for the machines nobody wants to shut down.
Why safety professionals stop going
It is not laziness. It is drift. The job accumulates administrative weight — recordkeeping, training records, corrective action tracking, reporting, audits, incident paperwork. All of it is real work, all of it is required, and all of it can be done sitting down. A safety manager can be genuinely busy for a year without spending meaningful time where the hazards are.
Then there is the social cost. Walking the floor means having conversations you would rather not have. Telling a supervisor his crew is cutting a corner. Being told you do not understand how the job actually works. Finding something that means a line goes down. Sitting at a desk has none of that friction, and friction is what people avoid.
What the walk actually finds
In my experience the recurring findings cluster in the same places every time:
Defeated controls with a rational explanation. A guard is removed because the setup cannot be performed with it in place. An interlock is bypassed because the machine cannot be threaded otherwise. These are not discipline problems — they are design problems, and you cannot find them without watching the task performed.
Work that does not match the procedure. The written procedure describes six steps. The operator does four, because two were written by someone who never ran the machine. Either the procedure is wrong or the practice is, and the gap is your citation.
Off-shift reality. Programs are written around first shift with full staffing and supervision. Second and third shift run leaner, with different informal norms. If you have never walked at 2am, you have assessed half your operation.
The thing nobody reports. Employees will tell you things in the aisle that they will never put in a report. That conversation is the highest-yield diagnostic tool in the profession and it costs nothing.
How to actually do it
Go with no agenda sometimes. A structured audit finds what the checklist looks for. An unstructured walk finds what you were not looking for, which is usually what matters.
Watch the task, not the area. Standing in a department tells you about housekeeping. Watching a changeover, a cleanout or a maintenance job tells you about exposure.
Ask open questions and then shut up. What is the most annoying part of this job? What would you change if you could? When was the last time something almost went wrong? Then let the silence do the work.
Close something visibly. The first walk generates credibility only if something you found gets fixed and people see it. Otherwise you have confirmed that walking around produces nothing.
Vary the time. Same route, same time, same day produces a performance rather than an observation.
This is the part that does not scale
I built a practice on automating the parts of EHS that should be automated. Document production, cross-referencing, gap mapping — that work is faster and cheaper done by machine with a credentialed professional reviewing it. That is exactly why the remaining human hours are so valuable, and it is why I spend them on the floor rather than at a keyboard.
If you are paying someone to own safety at your operation, ask them a simple question: how many hours did you spend on the production floor last month, and what did you find? If the answer is uncomfortable, that is your answer.
Key takeaways
- Documents are claims, not facts. Every program, training record and inspection log asserts something about reality that only a walk can verify.
- Drift is the cause, not laziness. Administrative load and social friction both pull the safety role toward the desk.
- Watch tasks, not areas. Changeovers, cleanouts and maintenance jobs carry the exposure that a departmental walkthrough misses.
- Off-shift is half your operation. Programs written around first shift describe conditions that do not exist at 2am.
- Close something visible. Credibility comes from a finding that gets fixed where people can see it, not from the walk itself.
Related reading: Fractional Safety Officer · Safety Culture Assessment · Incident Investigation & Corrective Action
