Insights / What aviation safety practice teaches other operations
"Be careful" is not a corrective action
The core of aviation safety thinking is that "pay closer attention" is not accepted as a fix. People miss things when tired and skip steps when familiar. You design assuming that, and build mechanisms that do not rest on individual vigilance.
That assumption transfers directly to visual inspection, to rostering, to stock control. Any process running on "the experienced one can tell just by looking" is already producing lower quality on the days that person is away — the problem is simply invisible.
Three transferable mechanisms
| Mechanism | Role in aviation | Translation elsewhere |
|---|---|---|
| Checklist | Read the procedure aloud before departure; confirm from a list, not from memory | Inspection items, shift handover, opening procedures written down and ticked |
| Redundancy | Duplicate systems so no single failure stops the operation | No task owned by exactly one person. A named backup and a written procedure |
| SOP | A standard procedure that holds quality steady whoever is on duty | Document the decision criteria so the veteran's instinct becomes repeatable |
Rostering, concretely
Operations control is the daily work of rearranging limited crew and aircraft while holding to safety standards and duty-time limits. Building a feasible assignment under constraints — that is structurally the same problem as building a staff roster.
What the exercise is really for is less about saving the hours spent drafting and more about translating the veteran scheduler's instinct into conditions you can write down. Once the constraints exist as text, the roster survives a change of personnel. How that looks in practice — plain-language rules going in, a month's roster coming out — is described in the guide to AI shift scheduling.
And inspection
Building inspection has the same shape. Where to look, what counts as a defect, what is tolerable until next year — while those live in one person's experience, quality varies with who is holding the clipboard.
Move the criteria into a model that returns the same output for the same input, and the variance disappears. What remains is a judgement call about what to do, which is where human expertise is actually worth paying for.
Who is writing this
KIYOMINE is the practice of Tsuyoshi Kiyomine, working on AI for operations that move — aviation, logistics and inspection. The common thread is replacing the step where a person has to be on site, look at something, and decide.
For how that reaches the field, see more about the practice.