Perspective · Culture & Strategy
Culture eats strategy. And culture runs on autonomy.
The phrase is usually attributed to Peter Drucker, though no published Drucker text contains it. Whoever said it first, the observation keeps being made: a strategy executed inside a culture that does not support it tends to be defeated by the culture, not by the strategy's internal logic. The interesting question is therefore not whether the phrase is true — it reliably is — but where culture actually comes from. Because if culture can be designed, the sequence reverses. If it cannot, strategy has to bend to whatever culture exists. Most organisations drift somewhere between those two positions, usually without noticing. What follows is a working theory of the shape culture actually sits on.
Peter Matysiak · Berlin · April 2026 · 9 min
A phrase that keeps being quoted.
"Culture eats strategy for breakfast" is usually credited to Peter Drucker. The attribution is uncertain — no published Drucker text contains the sentence — and the phrase was popularised by Mark Fields at Ford around 2006, credited to Drucker without a citation. Whoever coined it, the observation keeps being repeated because it keeps being confirmed: a strategy executed inside a culture that does not support it tends to be processed through the culture and come out the other side looking like whatever the culture already wanted to do.
The operator's question, then, is not whether the phrase is accurate. The interesting question is where culture comes from. Because if culture can be designed, the sequence reverses — the strategy is chosen to sit inside a culture that has been deliberately built. If culture is treated as fixed, the strategy must bend to match whatever culture already exists. Most organisations drift somewhere between those two positions, usually without being able to say which.
Three layers, only one of them visible.
Edgar Schein's three-level model (Organizational Culture and Leadership, 1985; now in its fifth edition) is still the cleanest account. Culture has three layers. Artifacts — what you can see: the office layout, the rituals, the dress code, the meeting patterns, the way people greet each other. Espoused values — what the organisation says about itself: the mission statement, the principles, the leadership memos. Basic underlying assumptions — what is actually believed: the unstated positions on how the world works, what people are like, what counts as success, what is safe to say.
The distance between the second and the third is where "culture eats strategy" happens. A strategy is always espoused — written in slides, rolled out in all-hands meetings, tied to objectives. Basic assumptions do the actual work. When the assumptions disagree with the strategy, the strategy is processed through the assumptions, and the output is whatever the assumptions were ready to produce anyway.
Schein is not alone in this reading. Chris Argyris distinguished "espoused theory" from "theory-in-use" in the 1970s and documented, in study after study, that the two diverge almost always — and that the divergence is invisible to the people inside it. Most published culture work since rests on this distinction, whether it credits Argyris or not.
What you do, is what it becomes.
Ben Horowitz makes the operator's version of this argument in What You Do Is Who You Are (2019). Culture is not what a leader says it is. Culture is what the organisation does repeatedly, especially when no one is watching. Every rule enforced or not enforced, every proposal rewarded or ignored, every poor decision punished or excused — each adds a layer of residue. Over years, the residue is the culture.
The practical consequence is that a leader has far less control over culture on the day of the announcement than on the thousand ordinary days that precede and follow it. Culture is slow; strategy is fast. When the two disagree, culture wins — because culture is the default setting the system reverts to the moment the strategy stops being enforced.
This is why the speeches rarely work. A speech changes espoused values. It does not change basic assumptions. Basic assumptions change through consequences — the observed response of the system to the thousand small tests it runs every day. A single visible consequence that contradicts the espoused value does more damage to the strategy than a hundred repetitions of the espoused value do good.
A shape the culture rests on.
One place the system registers more clearly than most is the pyramid of autonomy. The term has several close relatives in the management literature — Tannenbaum and Schmidt's leadership-behaviour continuum (HBR, 1958), Jurgen Appelo's delegation levels (Management 3.0, 2010), L. David Marquet's ladder of leadership (Turn the Ship Around!, 2012). They describe the same underlying shape in different numbers of rungs. At the base: "tell me what to do" — authority fully retained above. At the top: "I will inform you afterward" — authority fully placed with the person doing the work. Between those extremes, half a dozen intermediate rungs that shift the locus of decision step by step.
Every organisation sits on this pyramid somewhere. Very few sit on it coherently. Most spread across multiple rungs in different places, usually without the people involved being able to name which rung applies where. The inconsistency produces a predictable symptom: people play to the lowest observed rung, because that is the safest bet under ambiguity. The formal structure says level 5; the observed behaviour of the leader when a decision goes wrong says level 2; the team acts at level 2.
People play to the lowest observed rung, not the stated one. The stated rung is espoused value. The observed rung is basic assumption. Working note
Say the rung.
Jurgen Appelo's Delegation Poker is the cleanest operational expression I know. Seven levels: Tell, Sell, Consult, Agree, Advise, Inquire, Delegate — from authority fully retained to authority fully released. The move that works is naming the level for each class of decision, out loud, in the open, before the decision needs to be made. Not "use your judgement" — level 5 or level 6. Not "run it by me first" — level 3 or level 4. Not "decide and let me know" — level 7.
The effect of naming the level is disproportionate to its cost. It collapses the ambiguity that otherwise drives people to the base of the pyramid. It makes disagreements visible at the right time — before the decision, not after it. And it surfaces the inconsistencies: when a leader claims to be operating at level 5 but reliably acts at level 3, the gap becomes legible to everyone and can finally be discussed.
None of this is radical. The framework is fifteen years old and sits in ordinary agile-management training. What is radical is the discipline of doing it — of publicly naming the delegation level, sticking to it, and correcting it only with a stated change rather than a silent override. The cost is a few minutes per class of decision. The return is that the pyramid becomes legible instead of assumed.
Without it, the pyramid collapses.
The rungs above the base of the pyramid require one precondition that sits entirely in the culture: psychological safety. Amy Edmondson's long research programme at Harvard (The Fearless Organization, 2018) defines it as the stable belief that speaking up, disagreeing, admitting an error, or proposing something unpopular does not carry disproportionate personal risk. It is the cultural substrate that lets the pyramid function above its base rung.
Google's Project Aristotle (2015) is the most widely cited empirical test. A two-year study of 180+ teams across the company, looking for the factors that distinguished high-performing teams from low-performing ones. The dominant signal was psychological safety. Not talent composition. Not incentive design. Not process maturity. Psychological safety outperformed every other variable the researchers measured — by a wide margin, and consistently across very different team types.
The implication is strict. Without psychological safety, delegation above level 1 does not work, even when it is formally granted — people retreat to the base rung regardless, because the observable cost of being wrong at a higher rung is too high. The autonomy pyramid is a formal structure. Psychological safety is what lets the structure carry weight.
Design the strategy to fit the pyramid. Or change the pyramid first.
The operator's sequence follows directly. Before committing a strategy, diagnose the pyramid — the rungs people are actually operating on, not the rungs the org chart claims. If the strategy can be executed within the existing pyramid, execute it; no cultural work is required, and the rollout can be fast. If the strategy needs a different pyramid to work — more autonomy at the front, faster decision cycles, more disagreement in the open — then the pyramid has to be changed before or during the strategy, not after.
Changing the pyramid is slower than stating a new strategy, which is the part leaders find uncomfortable. It is also faster than watching the strategy be eaten by a culture that could not support it. The cost of skipping the diagnosis compounds across every quarter the strategy is nominally in effect and structurally not landing — and that cost usually gets booked as an execution problem, a talent problem, or a market problem, when it was a cultural feasibility problem from the start.
This is the part of the Drucker aphorism that does useful work. It is not a complaint about culture. It is a budget item. Strategy work carries a cultural prerequisite. If the prerequisite is not met, the strategy costs more than it looks like it costs — often by a multiple that is only visible in hindsight.
Where these arguments come from.
A short, opinionated reading list. Every one of these is worth reading in full. None of them should be read uncritically.
Talk it through?
A thirty-minute call usually surfaces the one pyramid-of-autonomy mismatch that is quietly defeating whichever strategy the team is running. The most useful calls come from leaders who want to test their strategy's cultural feasibility before spending another quarter on it — or who suspect the two are out of alignment and can name the symptom in a sentence.