Summary
AI can raise the apparent quality of HSE documentation while quietly lowering the competence that makes it meaningful.
We have always judged technical work partly by how it looks. AI breaks that shortcut: weak analysis can now look excellent.
The skills most at risk are observing, interpreting and challenging, and they fade without anyone noticing, because the documents keep looking good.
The better AI gets at producing professional output, the more competence you need to check it, not less.
In the first post of this series,
In the [first post of this series – The Competent HSE Professional, Amplified – I made the positive case. A competent HSE professional using AI well can outperform the same professional without it. But I attached a condition: the benefit belongs to the professional who stays in charge of the reasoning.
This post is about what happens when that condition is not met. Not in a single dramatic failure, but gradually, over months and years, in ways that are easy to miss.
In my previous series I described the false authority of AI output: how it can look complete, confident and correct when it is none of those things. That was a problem with the output. This is a problem with the people. What does working alongside false authority every day do to the professionals responsible for catching it?
(https://www.redlogenv.com/hse/the-competent-hse-professional-amplified), I made the positive case. A competent HSE professional using AI well can outperform the same professional without it. But I attached a condition: the benefit belongs to the professional who stays in charge of the reasoning.
This post is about what happens when that condition is not met. Not in a single dramatic failure, but gradually, over months and years, in ways that are easy to miss.
In my previous series on AI and HSE, I described the false authority of AI output: how it can look complete, confident and correct when it is none of those things. That was a problem with the output. This is a problem with the people. What does working alongside false authority every day do to the professionals responsible for catching it?
Competence is not knowing facts
When we talk about AI and skills, the conversation usually turns to knowledge. Will people remember the regulations? Will they know the exposure limits? That is the wrong worry. Facts can always be looked up, and AI is good at looking them up.
HSE competence is something else. It is a chain:
Observation → interpretation → challenge → judgement → action.
You notice something on site. You work out what it means. You question whether the plan still holds. You decide. You act, or you stop the job.
AI can help with parts of that chain. It cannot stand on the site for you. And the links most at risk are the first three, because they are exactly the ones a well-written document makes feel unnecessary.
Consider a simple, hypothetical example. A supervisor uses AI to prepare a job safety analysis for a lift. The document is excellent: well structured, the right terminology, every standard hazard covered, controls properly ranked. It would pass any desk review.
But overnight the ground beside an excavation was backfilled, and the crane is due to set up on it. The document cannot know that. The only safeguard is a person who walks the ground, notices it, and has the confidence to say the JSA no longer fits the job.
The question is whether that person is still developing the habit of looking, or whether they are developing the habit of trusting the document.
The fluency trap
For most of my career, poor technical work tended to look poor. A weak risk assessment was usually vague, incomplete, badly organised or full of generic phrases. Experienced reviewers learned to read those signals. Structure, precision and confident technical language were imperfect but useful clues to the quality of the thinking behind them.
AI breaks that link. A weak analysis can now be beautifully structured, professionally written, apparently comprehensive and confidently expressed. The surface signals we relied on, often without realising it, no longer tell us anything about the substance.
That is the fluency trap. When the form is always good, the reviewer must judge the substance directly, and that takes more competence than reading the form ever did.
It leads to what I think is the central management implication of AI in HSE:
The more capable AI becomes at producing professional output, the more important independent human competence becomes for checking it.
That sounds like a paradox. It is not. It is simply what happens when quality can no longer be judged by appearance.
Skills fade quietly
Skills that are not used decline. Every HSE professional knows this from emergency drills: the response that was sharp in a monthly exercise is slow after a year without one.
There is now early evidence that the same applies when AI does part of the work.
A study published in *The Lancet Gastroenterology & Hepatology* in 2025 looked at experienced endoscopists in Poland after AI assistance was introduced into routine colonoscopy. When those same doctors later performed procedures without AI, their detection rate for pre-cancerous growths fell from 28.4% to 22.4%. The study is observational, and the authors are careful to say other factors may have played a part. But it is the first real-world clinical evidence of what researchers call deskilling: routine AI assistance eroding the skills needed to perform well without it.
A 2025 survey by Microsoft Research and Carnegie Mellon University of 319 knowledge workers points the same way. The more confidence people had in AI’s ability to do a task, the less critical thinking they reported applying to it.
Neither study is about HSE. Both describe exactly the mechanism we should expect in HSE: when the tool is usually right, people stop checking, and the ability to check fades with the habit.
What makes this dangerous is that nobody notices. The documents keep looking good. The audits keep passing. The decline only becomes visible when something unusual happens and the person who should have spotted it does not. In HSE, that moment is often an incident.
The apprenticeship problem
There is a second, slower effect, and it worries me more.
Experienced professionals can use AI relatively safely because they already know what good looks like. They built that sense over years of doing the work themselves: drafting the risk assessments, reading the regulations, getting things wrong and being corrected by someone more senior.
Much of that formative work is exactly what AI now does fastest. If junior HSE professionals never draft a risk assessment from a blank page, never trace an obligation back to the regulation, and never have their reasoning challenged, where does their judgement come from?
The risk is not that today’s experts become incompetent overnight. It is that the next generation never becomes expert at all, while the documents they produce look as good as their seniors’ did.
Why HSE is unusually exposed
In many fields, a fluent but shallow document costs time or money. In HSE, the activities behind the documents include confined-space entry, lifting operations, process isolation, chemical exposure, electrical work, work at height, excavation and emergency response.
Compliance, too, is a chain: from requirement to obligation, operational control, evidence, verification and corrective action. AI can assist every link. But somebody still needs enough subject knowledge to recognise when the chain is broken, and a polished document can hide a broken link very effectively.
Could your people still do it without the tool?
This is the question I would put to every HSE leader. We will explore it properly in the webinar, but a few practical starting points are worth considering now.
Test reasoning, not documents. Competence assessments that check whether paperwork is complete will not detect deskilling. Ask people to explain and defend their decisions, on site where possible.
Keep some work tool-free. Site walk-downs, verbal challenge of a risk assessment, and investigation interviews all build the skills AI cannot exercise for you.
Protect the learning tasks. Decide deliberately which work junior staff should do themselves first, before AI is allowed to help.
Look at what changes at review. If AI output routinely passes unchanged, find out whether that is because it is right or because nobody is really looking.
From the individual to the system
The fluency trap is a problem for individuals. In the next post I look at what happens when it scales across an organisation: when the worker, the supervisor, the HSE manager and the auditor all check the work with the same AI. You may think you have four lines of defence. You may only have one.
Why Redlog Is Building for This Future
At Redlog, we are building compliance intelligence tools on the principles set out in my previous series: curated and expert-validated regulatory content, jurisdiction-specific coverage, source traceability, applicability-aware outputs, and an expert-in-the-loop model that keeps professional accountability where it belongs.
This series adds a further principle. AI should strengthen the professional’s judgement, not stand in for it. Tools that show their sources and flag their uncertainty prompt the professional to check, challenge and decide. That keeps them in the left-hand column.
We are publishing this series because the conversation about AI and accountability in HSE should happen inside the profession. It should be shaped by people who understand regulatory compliance, duty of care, operational risk and the consequences of getting this work wrong.
AI will change HSE practice. That is no longer the question. The question is whether the profession will shape that change with judgement, governance and accountability, or allow it to be shaped elsewhere.
If that is how you think AI should work in HSE compliance, it is the conversation we want to have, starting with our webinar.
This is Part 2 of AI and HSE Management: The Good, the Bad and the Accountable. The previous series, AI and the HSSE Profession: What’s Actually Changing, and all of our prior blog posts are available at www.redlogenv.com/blog.
Webinar – Join us on 10 November 2026
Webinar: AI, Accountability and HSE Compliance in MENA — What Every Business Leader Needs to Know
Join us for a discussion with our experts to learn more about where liability sits when AI-assisted compliance work goes wrong, what regulators are starting to expect, and what good governance looks like in practice, and more. Register here
AI and HSE Management: The Good, the Bad and the Accountable — Part 2 of 8
- The Fluency Trap - October 6, 2026
- The Competent HSE Professional, Amplified - September 29, 2026
- Confused About Heat Stress and the GCC Midday Work Ban? - June 27, 2026