The core risk
The AI replacement narrative assumes expertise can be extracted from humans, embedded into software, and scaled while the human labor system is reduced. In safety-critical work, that assumption is structurally dangerous.
AI systems depend on continuing contact with operational reality. Without new, grounded, human-generated information, systems can narrow toward averages, lose rare cases, and erase low-frequency signals.
Why this is an operational safety issue
Safety work depends on weak, contextual, and tacit signals: the unusual vibration, the informal warning, the missing context, the edge case that does not fit the dashboard. These are often the signals that prevent harm.
If organizations remove the people and roles that generate, challenge, correct, and ground those signals, they degrade the same knowledge system AI depends on.
HAOP framing
HAOP treats this as an epistemic risk: the work system can lose its ability to know what is actually happening. Grounding is not a model feature alone. It is a work-system control that must preserve human expertise, feedback loops, verification, and contact with real conditions.