A Method Built on Disciplined Questions
HAZOP's staying power comes from its deceptive simplicity: apply the same small set of guide words — more, less, none, reverse, as well as — to every node of a process, systematically and without exception, with the right people gathered in the room at the same time. That discipline is precisely what catches the deviations checklist-based audits routinely miss, because a checklist can only ever ask about failure modes someone already thought to write down in advance.
What's changed over the years isn't the method itself so much as the depth teams now bring to each session. The best HAZOPs today pull in real operating data, historical deviations, and near-miss reports, rather than relying purely on the original design intent captured in old process drawings that may no longer reflect how the unit actually runs.
Where AI Genuinely Helps
Large language models are useful for surfacing candidate deviations from historical incident databases, drafting node descriptions, and speeding up documentation — tasks that used to eat hours of a facilitator's time without adding much judgement value.
But consequence severity, likelihood ranking, and reading the room when an operator hesitates before answering a question are still fundamentally human skills. AI can prepare the ground; it can't yet sit at the head of the table.
