On the recording from the Miami overrun, a pilot says the aircraft is going too fast about a hundred seconds before it crashes, and says it again. The go-around is called at the end.
No finding has been made and none of this is about that crew, whose account has only just been taken. But the shape of it — a problem identified aloud, early, by someone in the room, and a course that did not change — is the single most studied failure in commercial aviation, and it is one almost no other industry has done the work on.
Why aviation had to build machinery for this
The uncomfortable finding from decades of accident investigation was not that crews failed to notice things. It was that they noticed, mentioned it, and were not acted upon — and that this happened disproportionately when the person noticing was junior to the person flying.
The response was crew resource management, and the important thing about it is what it did not try to do. It did not attempt to make people braver, or to train assertiveness in the abstract, or to flatten a hierarchy that exists for good operational reasons.
It changed the speech act. It defined specific callouts with specific wording, made a response to them mandatory rather than discretionary, and gave any crew member an explicit duty — not a right, a duty — to demand a go-around, with the expectation that it be flown without discussion.
What most organisations built instead
The reporting half, and almost never the response half.
There is a channel. There is a form, or a retro, or a risk register, or a speak-up line. Someone raises a concern and the concern is recorded. What is almost never specified is what has to happen next: who must respond, within what period, and what they are obliged to do rather than free to do.
The result is a system that converts a warning into a ticket. This desk has described the same failure from several angles — complaints routed to resolution rather than to comprehension, and dashboards nobody reads because no decision hangs on them. In each case the information arrives and nothing is required of anybody on receipt.
The three properties that make a raised concern actionable
They are not complicated and they are almost never all present.
Named form. A concern has to be sayable in a way that is unmistakably a stop signal rather than a remark. "I'm not comfortable with this" is ambiguous by design, which is why the person saying it chooses it. Organisations that work at this give people a phrase that cannot be heard as conversation.
Mandatory response. Somebody must answer, and the answer must be a decision rather than an acknowledgement. A concern that can be received without a decision will be.
Protected escalation. If the answer is no, there must be somewhere else to go that does not require the objector to be right. The cost of being wrong has to be near zero, because otherwise only certainties get raised, and certainties are rare early.
The measurement that would show whether it works
Not the number of concerns raised, which every organisation counts and which mostly measures how much people trust the channel.
The interval between a concern being raised and a decision being recorded against it. That single number describes whether the system responds or merely receives, and almost nobody instruments it — which is itself informative, because it is the easiest thing in the process to measure.
And the honest caveat
The interval on that recording may turn out to have an explanation that has nothing to do with any of this. Crews get behind aircraft for many reasons, the investigation is days old, and the pilots have only just been interviewed.
The reason to write about it now is not to explain a crash. It is that most organisations reading this have a hundred-second gap of their own somewhere, and no way of knowing, because they count the warnings and not the time it takes to answer one.
The cockpit voice recorder finding that a pilot warned of excessive speed about one minute and forty-two seconds before the crash and continued to comment on speed, and that a go-around was called seconds before impact, is as reported by ABC News and CNN on 9 and 10 September 2026 and covered in this publication's news report. This is a column. No finding has been made by the NTSB, no inference is drawn about the conduct of this crew, and the general account of crew resource management practice is not specific to this operator. The organisational argument is the author's own.
Topics leadershipmanagementsafetycommunicationdecision making
