An anchor is a short sentence, chosen in advance and repeated aloud, that names what you protect when a protocol, a metric or a system tells you to do something else. Change readiness is usually described as flexibility. Flexibility with nothing fixed at the centre is drift with a better name.
The answer, in one line
A short sentence a team chooses in advance and repeats, naming what is protected when the system says otherwise. Protect the patient in front of you is an example from the book.
Definition#
The anchor principle: a team's short, repeated sentence naming what is protected when the system directs otherwise. Stillness at the centre is what makes rapid adaptation safe rather than arbitrary.
Why an anchor has to exist beforehand#
An anchor only does work at the moment it puts you at odds with something. That moment is always under time pressure, usually late, and frequently in front of somebody who outranks you. It is the one moment at which the sentence cannot be composed, because whatever you compose then will be shaped by what is convenient.
So the anchor is written in advance, by the team rather than for it, in words short enough to be said out loud at a shift change. The book's example is a line above a nurses' station: protect the patient in front of you. Six words, chosen months before the night they mattered.
Anchors cost something#
This is the part usually left out of the values conversation. The book's account of the nurse who ordered a test the protocol did not require ends with the patient stabilised and the nurse reprimanded for delaying throughput. Her record showed deviation. Her conscience showed alignment.
A values statement that has never cost anybody anything has not been tested and probably does not exist. An anchor is written expecting the bill, and the organisational question is whether the bill falls entirely on the individual who held the line. Where it does, the anchor survives about as long as the person does.
What this has to do with machines#
Automated systems supply instructions quickly and with apparent confidence, and the instruction usually arrives with a number attached. A person deviating from it needs something to weigh it against, or the deviation registers only as error.
That makes the anchor the sentence a person reaches for on override terrain, where the cost of being wrong falls on someone rather than on a system. It is also what makes the deviation legible afterwards. Without an anchor, a professional who overrode a system has a judgement call and no vocabulary for it, which is the position described at who can override an AI system.
Writing one#
- Short enough to say out loud without reading it.
- Specific enough to conflict with something. If no protocol could ever contradict it, it is a slogan.
- Chosen by the people who will have to hold it, not issued to them.
- Repeated at the moments when nothing is happening, which is what makes it available when something is.
What this has not been shown to do#
No trial exists. Nothing here shows that teams with anchors deviate better, or that the anchor rather than the individual's own judgement produced the deviation in any recorded case. The account is argued from practice, and the honest reading is that an anchor gives a person language for a call they might have made anyway, and makes it defensible afterwards to people who were not there.
Source#
Rahim Hirji, SuperSkills: The Seven Human Skills for the Age of AI, Kogan Page, 2026, chapter two. The cases are in the book.
Related SuperSkills research#
Essay · SS-2026-336
Hirji, R. (2026). The anchor principle. The SuperSkills evidence base, SS-2026-336. https://thesuperskills.com/research/the-anchor-principle. Last reviewed 26 September 2026.
An evidence review by Rahim Hirji, not peer-reviewed research. For a material claim, cite the underlying study as well; every study here carries its own permanent link.
How citations and IDs work