The mechanism sounds minor: change “request permission to do X” to “I intend to do X.” The person receiving the communication can still stop it — nothing prevents a senior officer from saying no. The content is essentially the same. What changes is where the cognitive ownership sits.
In the request-permission formulation, the thinking is with the person who grants or denies. They must assess whether to allow the action. The person making the request has proposed something and is waiting. In the “I intend to”formulation, the thinking is with the person who will execute. They have decided, stated their decision, and the person above them has the opportunity to intervene — but defaults to trust rather than to checking.
This is Marquet’s account from Turn the Ship Around!, drawn from his command of the USS Santa Fe. The mechanism is one of the most precise descriptions of how distributed authority is operationalised in practice — not as a policy but as a linguistic habit that shifts cognitive ownership.
What changes when language changes
The person who says “I intend to” has done something the person who says “request permission” has not: they have committed to a direction. They must think the action through well enough to state it. They are not waiting for someone else to confirm that it is correct — they are announcing their own assessment and inviting correction rather than approval.
Over time, this shifts the distribution of thinking in the organisation. More people are doing the reasoning; fewer people are waiting for someone else to do it. The organisation runs at the speed of the people closest to the work rather than at the speed of the approval chain.
The transfer from a nuclear submarine to other workplaces requires a small translation but not a large one. Consider two versions of the same message in a project environment. First: “Should we proceed with the supplier review before the board meeting?” Second: “I intend to run the supplier review before the board meeting — let me know if there are reasons not to.” The content is nearly identical. The first invites the recipient to make the decision. The second places the decision with the sender and invites a veto. Everyone who reads the second message understands immediately that this person has thought it through.
The preconditions — and the failure mode
The preconditions Marquet identifies still apply: this mechanism works when the people using it have the competence to make the decisions they are announcing and the clarity about purpose to make decisions that cohere. Without competence, “I intend to” distributes errors. Without clarity, it distributes decisions that pull in different directions. With both, it distributes capability.
The failure mode is worth naming: an organisation that adopts the language without building the preconditions gets confident, fast, local errors — and no approval chain left to catch them. “I intend to” is not a cultural shorthand for autonomy. It is the operational expression of authority that has been genuinely earned and clearly bounded. The mechanism only works as advertised when both of those conditions are true.
What “request permission” signals over time
There is also a slow cost on the other side that is easy to miss. An organisation in which people routinely ask permission rather than state intent is not just slow — it is training people not to think. The request-permission habit, repeated across thousands of small decisions over months and years, teaches people that their judgement is not trusted and not theirs to exercise. They learn, accurately, that the thinking is supposed to happen somewhere above them. Eventually it does — only there, and less and less anywhere else.
Turn the Ship Around! (2012, Portfolio/Penguin) by L. David Marquet. Single-case account from one naval command; transfer to other settings requires judgement about context.
The structural principle behind this mechanism is developed in Move Authority to the Information. The paired military-leadership comparison is in Decentralised command: the canonical pair (in development).