Organisations spend heavily to teach leaders what good leadership looks like, and the teaching largely works: leaders can articulate the models with fluency. Performance, however, moves very little. This paper treats that failure to transfer — the knowing–doing gap — as the central problem of leadership development rather than an unfortunate side effect of it. It argues that the dominant skills model cannot account for the gap, because the gap is not a deficit of knowledge but a feature of where action actually comes from. Drawing on the ontological tradition in cognition and language, it offers a three-layer model — Knowing, Being, Doing — that locates the gap precisely, explains why more knowledge often widens rather than closes it, and specifies what an intervention would have to do to produce change that holds under pressure.
Every serious organisation already knows what good leadership looks like. It has the competency frameworks, the 360 instruments, the reading lists, the off-sites. Its leaders can describe, with real fluency, what they ought to do: listen first, set clear expectations, give candid feedback, hold people to account, make space for dissent. Ask them in the room and they will tell you. Then watch them on a Tuesday afternoon under load, and a different person shows up — faster, more controlling, reaching for the familiar move rather than the right one.
This is not hypocrisy and it is not a motivation problem. It is the most reliable finding in the whole field, and it has a name. Jeffrey Pfeffer and Robert Sutton called it the knowing–doing gap: the persistent space between what people in organisations know and what they actually do. They documented it as the difference between companies that possess knowledge and companies that act on it, and they were unsparing about how common the failure is. Two decades of leadership development since has not closed it. If anything, the industry has gotten better at the part that does not work — the transfer of knowledge — while leaving the part that matters largely untouched.
The argument of this paper is simple to state and uncomfortable to sit with. The knowing–doing gap is not a gap in knowing. It is evidence that knowing was never the thing that produced doing in the first place. We have built an entire development apparatus on the assumption that better-informed leaders lead better, and the assumption is false often enough to be treated as the exception's exception. To close the gap we have to stop trying to close it with more knowing, and look instead at the layer underneath performance that the skills model never touches.
Consider a composite anyone who has run a leadership programme will recognise. A capable executive — on any reckoning the most coachable person in the cohort — completes a programme on inclusive leadership. She can define psychological safety, cite the studies, and explain with conviction why a leader's first job in a meeting is to make it safe to disagree. Her knowledge is, by any test, excellent. Six weeks later her team's engagement has not moved, and in the room she is exactly who she was before: quick, decisive, faintly impatient with the slow formation of other people's thoughts. Nothing failed in her learning. Something simply never crossed from what she learned to who she is when a meeting is running late and a junior colleague is taking too long to get to the point. The skills model has no account of that crossing, because in its picture the crossing is automatic and instantaneous. The entire purpose of this paper is to describe the terrain that crossing actually has to cover — and why, for most leaders, it never gets crossed.
The stakes of getting this wrong are not academic. Leadership development is among the largest discretionary investments organisations make in their own capability, and the public-sector and corporate functions that commission it are increasingly asked to show a return. When the return does not materialise — when leaders emerge from expensive programmes able to describe better leadership without exercising it — the conclusion drawn is usually that the wrong programme was bought, and a different one is bought next year on the same underlying model. The gap is not a procurement error. It is structural, and it will reproduce under any programme built on the assumption that knowing leads to doing.
For most of its history, leadership development has run on what we can call the skills model. Its logic is clean and intuitive: leadership is a set of competencies; competencies can be specified, taught, and assessed; therefore a leader who acquires the competencies will lead more effectively. Almost every artefact of the modern development function — the framework, the diagnostic, the curriculum, the individual development plan — is an expression of this single idea.
The skills model makes a testable prediction: increase a leader's leadership knowledge and you should see a corresponding increase in leadership behaviour. This is precisely where it fails. The relationship between what managers know about leadership and what they do as leaders is weak, and recent research has put the failure on a firmer footing. A study published in the Academy of Management Learning & Education, taking a phenomenological approach to the experience of managers across organisational levels in the United Kingdom, defined the leadership knowing–doing gap as the discrepancy between what managers know about leadership and the extent to which they transfer it into leadership doing — and found it to be a multifaceted, dynamic experience with cognitive, affective, and behavioural elements, not a simple shortfall of information.
More striking still is a finding that ought to be fatal to the skills model: in many cases, more leadership knowledge makes the gap worse, not better. A leader with a richer conceptual vocabulary has more sophisticated ways to describe what they failed to do, more frameworks to retreat into, more language to rationalise the gap after the fact. The knowledge does not convert into action; it converts into a more articulate account of inaction. Any model in which the central input — knowledge — is sometimes negatively related to the desired output is not a model with a tuning problem. It is a model pointed at the wrong variable.
We have built an entire development apparatus on the assumption that better-informed leaders lead better — and the assumption is false often enough to be treated as a rule.
It is worth being precise about what the skills model gets right, because the answer is not "nothing." Knowledge is necessary. A leader who has never encountered the idea of psychological safety, or never learned to separate an observation from an evaluation, is missing raw material they cannot act on. The error is not in valuing knowledge. The error is in treating knowledge as sufficient — in believing that the journey from understanding to action is short, automatic, and frictionless, when in fact it crosses the most consequential terrain in the whole of leadership, terrain the skills model does not even have a map for.
To see the gap clearly we have to ask a question the skills model never asks: where does action come from? Not in theory — in the moment. When a leader does something effective under pressure, what produced it?
It was not, in the live instant, a piece of knowledge being retrieved and applied. Watch anyone perform at a high level in any domain and the knowledge has disappeared into the performance. The surgeon is not recalling anatomy; the anatomy is using her. The point guard is not remembering the playbook; the game is playing through him. The philosopher Hubert Dreyfus spent a career arguing exactly this against the cognitivist account of expertise: genuine skill is not the rapid application of rules but a way of being absorbed in a situation, responding to its demands without the intermediate step of deliberation. The expert has stopped consulting a model and started dwelling in the domain. The rules were scaffolding; they fell away when mastery arrived.
This is the distinction the ontological tradition draws between knowing about a world and being in it — and it is the hinge of this entire paper. Erhard, Jensen and their colleagues, in their ontological model of leadership, make the contrast vivid with a thought experiment borrowed from philosophy. Imagine a brilliant scientist who has learned everything that can be known about colour — every wavelength, every account of what each colour looks like — while living her whole life in a room without any colour in it. She has a complete conceptual grasp of colour and no experience of it whatsoever. The day she walks out into a world of colour, she gains something her vast knowledge never gave her. A conceptual grasp leaves you knowing. Being in the world leaves you being. The difference is not academic; it is the difference between describing tennis from the stands and playing it on the court.
Bring that distinction back to leadership and the knowing–doing gap stops being mysterious. A leader who has only studied leadership is in the stands. They can narrate the game with authority and remain, in any live moment, unable to play it — not because they lack information, but because information is not where playing comes from. The whole of conventional development is an attempt to make better spectators, and then a puzzled inquiry into why the spectators do not perform like players.
Before building the model it is worth naming the gap in the forms practitioners actually meet it, because it does not present as one thing. It presents as three, and each points at a different layer of what follows.
The first is the articulate non-doer. This is the leader whose language has improved and whose behaviour has not. They have absorbed the vocabulary — coaching, candour, empowerment — and deploy it fluently in describing both what good leadership requires and, after the fact, why circumstances prevented it this time. The fluency is real and it is also the trap: it gives everyone, including the leader, the impression that development has occurred, because the most visible output of development — better talk — is present in abundance. Knowledge went in; description came out; action did not change. This is the gap in its purest form, and it is the form the skills model is least able to see, because the skills model measures exactly the thing — articulate knowledge — that the articulate non-doer has in surplus.
The second is the pressure reverter. This leader can do the thing — genuinely — when conditions are calm. In a low-stakes coaching conversation they listen beautifully; in a workshop role-play they make clean requests. Then the quarter turns, a number slips, and within a week they are interrupting, deciding alone, and managing by exception. They have not forgotten what they know. The knowledge has become unreachable under the only conditions that matter, because pressure changed who they are being and sealed the knowledge off behind it. The pressure reverter is proof that the gap is not about knowledge at all: the same person, with the same knowledge, closes the gap when relaxed and opens it under load.
The third is the insight that never coordinates. This leader has done the inner work — has genuinely shifted how they see, become a more spacious and perceptive observer — and still cannot get a team to move. The realisation is real and it stays trapped inside them, because they cannot translate a new way of being into the concrete conversational moves that organise other people's action. They are wiser and no more effective. This face of the gap is the one most often mistaken for success, because the leader reports a profound personal change and means it — and the organisation sees no difference in what gets done.
Three faces, three locations. The articulate non-doer is stuck at the boundary between knowing and being. The pressure reverter is governed by a layer of being that overrides what they know. The insight that never coordinates has the being and lacks the doing. A model adequate to the gap has to hold all three — which means it needs at least these three layers.
If knowledge is not where doing comes from, we need an account of what is. The model below names three layers between a leader and their action, and locates the knowing–doing gap precisely — not as a missing rung on a ladder, but as an unbridged space between two layers that operate by entirely different logics.
This is the epistemic layer: everything a leader knows about leadership. Concepts, distinctions, models, the contents of every course they have attended. It is real and it matters, and it is also the only layer the development industry reliably operates on. Almost all the money, time, and design ingenuity in leadership development is spent enriching layer one. The trouble is that layer one does not, on its own, reach the floor where action happens.
This is the ontological layer, and it is the one the skills model has no concept of. It is not what a leader knows but who they are being — the observer they are constituted as in a given moment. Humberto Maturana's central claim, from the biology of cognition, is that everything said is said by an observer, and that what an observer can distinguish depends on the kind of observer they are. Applied to leadership, this is not mysticism; it is mechanism. The observer a leader is being determines what even shows up to them as present in a situation: which signals they notice, what they read a silence to mean, whether a hard conversation registers as a threat to manage or an opening to step into. Two leaders with identical knowledge, being different observers, inhabit different situations and therefore have different actions available to them. Layer two governs whether anything in layer one is ever reachable when it counts.
This is the coordinative layer, where being becomes organisational result — and, following Fernando Flores, it happens in language. Organisations do not run on tasks; they run on a network of commitments built and maintained in conversation: requests with explicit conditions of satisfaction, promises that are genuinely made, declarations that close work as complete. Action, at the scale of an organisation, is a sequence of these linguistic moves. A leader can be a different observer and still produce nothing if they cannot make a clean request or secure a real promise. Layer three is where the inner shift either coordinates other people or dissipates.
Two leaders with identical knowledge, being different observers, inhabit different situations — and therefore have different actions available to them.
Read top to bottom, the model tells a different story about performance than the skills model does. Knowledge enters at layer one. Whether it ever becomes action depends on layer two: the observer the leader is being has to be one for whom that knowledge is live and reachable, not filed and inert. And whether that reachable action becomes a result depends on layer three: the coordinative moves that turn a leader's intention into other people's coordinated effort. Performance is the whole stack working together. The knowing–doing gap is what opens when an organisation develops layer one and leaves layers two and three to chance.
It helps to run a single situation through all three layers. A director needs a struggling project brought back on track and has to address it with the manager responsible. At layer one she knows everything required: she can describe how to give difficult feedback, how to separate the behaviour from the person, how to agree clear next steps. None of that knowledge is in doubt. At layer two, the question is who she is being as she walks into the room — and on a bad day she is being someone for whom this conversation is a threat to be contained, which means what shows up to her is a defensive manager to be managed rather than a capable colleague to be levelled with. From inside that observer, the feedback she knows how to give is not available; what is available is a softened, hedged version that protects everyone from discomfort and changes nothing. Suppose, instead, she has done the work at layer two and walks in as someone for whom the conversation is an act of respect owed to a colleague. Now the same knowledge is reachable, because the situation she inhabits is one in which candour is the natural move. But she is still not finished. At layer three, she has to actually make the request — to name what specifically needs to be true by when, secure a real promise rather than a nod, and arrange the moment at which completion will be declared. Skip that layer and even the well-grounded conversation dissolves into mutual goodwill and no changed outcome. Only when all three layers hold does the project actually move — and the same is true, multiplied across a thousand such moments, of an organisation's whole performance.
The model also explains the two features of the gap that most embarrass the skills account: that it appears under pressure, and that more knowledge sometimes makes it worse.
Pressure is the load test of the whole stack, which is why the gap is invisible in the classroom and obvious on the Tuesday afternoon. In a low-stakes setting, a leader has time to consult layer one deliberately — to remember the framework and apply it. Pressure removes that time. Under load, a leader does not reach for what they know; they express who they are being. If the observer they collapse into under stress is a controlling, threat-managing one, then all the knowledge in layer one is sealed off behind a layer-two state that cannot reach it. This is why the leader who can describe great listening in the morning interrupts everyone by mid-afternoon. The knowledge did not leave. The conditions changed, layer two shifted, and the knowledge became unreachable. Development that only addresses layer one is, almost by definition, training for conditions that never test it.
The second feature — that more knowledge can deepen the gap — follows from the same structure. New knowledge added to layer one, with no change at layer two, does not give a leader new action. It gives them new description: more precise language for the gap, more sophisticated frameworks to retreat into, a richer account of why the right move was not available. Knowledge with no ontological change becomes commentary on inaction. The leader who has read the most about vulnerability can be the one most fluent at explaining, afterwards, why this particular situation was not safe enough for it. The skills model reads this as a need for still more knowledge. The three-layer model reads it correctly: as knowledge piling up against a layer-two wall.
If the gap opens between layer one and the layers beneath it, then closing it cannot be done from layer one. No amount of additional knowing reaches across the gap, because the gap is not made of missing knowledge. Closing it requires intervention where the skills model never goes: at being and at doing.
Working at layer two means working on the observer the leader is — not adding to what they know, but altering what they can perceive and what therefore becomes possible for them to do. This is the part that sounds soft and is in fact the most rigorous, because it cannot be transmitted as content. An observer cannot be told into existence; it has to be occasioned. The leader has to encounter, in their own direct experience, the observer they have been operating as — usually invisible to them, the way water is invisible to a fish — and discover that it is not the only one available. That discovery, when it is genuine and first-hand rather than received as another concept, is what lets a different range of action show up. It is also why this work is experiential by necessity, not by preference. The moment it becomes content to be learned, it has fallen back into layer one and lost its power.
This is also the point at which most organisations flinch, and the flinch is worth naming. Layer-two work is harder to specify in a procurement document, harder to schedule into a half-day, and harder to evidence with a satisfaction score than a content module is. It asks participants to confront something about how they have been operating, which is uncomfortable in a way that absorbing a framework is not. Faced with that, the path of least resistance is to retreat to layer one, where everything is legible and nothing is threatening — and to accept the knowing–doing gap as the price. The discomfort, however, is not a defect of the method; it is the method working. An observer that could be shifted painlessly was not the observer running the show. The work has to reach the place where the leader has genuinely been operating, and that place does not give way to comfort.
There is a further reason layer-two work resists the classroom: an observer cannot be argued into change, only occasioned into it. You cannot tell a leader that they have been seeing their team as a problem to be managed and expect the telling to dissolve it — the telling lands as one more concept, filed alongside the others. What dissolves it is an experience, constructed carefully, in which the leader catches themselves in the act of being that observer and sees, for themselves and not on anyone's authority, that it is a stance they have been taking rather than the way things simply are. That seeing is the hinge. Everything before it is preparation and everything after it is consolidation, and it cannot be delivered as information because information is the very register it has to break out of.
Working at layer three means making the coordinative moves explicit and trainable: how to make a request that actually specifies what would satisfy it, how to secure a promise rather than extract a vague assent, how to declare completion so that work is genuinely finished rather than left ambiguously open. These are not personality traits; they are competences in language, and they can be practised. Most of what an organisation experiences as an execution problem is, on inspection, a broken move at this layer — a request never clearly made, a decline never voiced, a completion never declared — and naming the broken move is the first step to repairing it.
The two work together. A shift at layer two without layer three is insight that never coordinates anyone; a leader becomes a more capable observer and still cannot get a team to move. Layer three without layer two is technique that collapses under pressure; the leader knows the moves and reverts to type the moment the stakes rise. Effective development holds both: it shifts who the leader is being and trains how that shows up in language. That combination is what produces change that survives contact with a real Tuesday.
A research desk owes more than a model; it owes a way to know whether the model is working. The knowing–doing gap has felt unmeasurable largely because it has been mislocated — treated as an attitude or a competency score rather than as a difference between what a leader knows and what they observably do. Located properly, it becomes assessable, because layers two and three throw off observable markers in a way that layer one does not.
At layer three, the markers are concrete and countable. The proportion of requests made with explicit conditions of satisfaction and a due time. The rate of commitments that are explicitly declined rather than silently dropped. The share of work that ends in a declared completion rather than trailing off unacknowledged. The volume of rework traceable to a request that was never clearly made. None of these require a survey; they are visible in how a team actually coordinates, and they move when the work is landing.
At layer two, the markers are subtler but still real: the range of situations a leader can stay resourceful in before collapsing into their default observer; the speed with which they recover after they do; whether, under pressure, their behaviour still tracks what they know or detaches from it. These can be observed over time and triangulated against the layer-three numbers. When the inner shift is real, the coordinative markers move with it. When training has only touched layer one, the leader scores well on a knowledge check and the coordinative markers do not move at all — which is, precisely, the knowing–doing gap, now visible as a measurement rather than a complaint.
When training has only touched what a leader knows, they pass the knowledge check and nothing they do actually changes. That is the gap, made visible as a measurement.
Taken seriously, the three-layer model reorders the priorities of the entire development function. If layer one is necessary but not where the gap lives, then the marginal hour and the marginal pound are badly misallocated when they go, as they overwhelmingly do, to enriching what leaders know. The leverage is at layers two and three, which is exactly where conventional programmes spend least.
This does not mean abandoning content. It means demoting it to its actual role — necessary raw material — and refusing to mistake its delivery for development. A programme designed around the model would spend its content budget efficiently and then spend the bulk of its time on the two things content cannot do: occasioning a genuine shift in the observer a leader is being, and drilling the coordinative moves until they hold under load. It would judge itself not by what participants can now articulate but by what observably changes in how they coordinate when no one is watching for it.
It also reframes what an organisation should buy. The right question to ask of any leadership intervention is not "what will our leaders learn?" but "what will our leaders be able to do under pressure that they cannot do now, and how will we see it?" An intervention that can only answer the first question is operating entirely in layer one and will, with high reliability, produce the knowing–doing gap it was bought to close. An intervention that can answer the second is working where the gap actually lives.
Two reasonable objections are worth meeting directly, because a model that cannot withstand them is not worth building a practice on.
The first: isn't this just a restatement that leaders need experience? Experience and time on the job clearly narrow the gap for some, after all. But experience is not a mechanism; it is a container in which a mechanism may or may not operate. Plenty of leaders accumulate decades of experience and never close the gap — they get more practised at their default observer rather than freer of it, which is why seniority and the knowing–doing gap so often coexist. What experience sometimes does, when it does anything, is occasion the layer-two shift this paper describes: a failure painful enough, or a moment exposed enough, that a leader catches their own stance and it loosens. The model does not deny that experience can produce the shift. It denies that experience reliably does, and proposes that what experience occasions by accident can be occasioned on purpose, faster and for more people, by working directly where experience occasionally lands rather than waiting for it to land.
The second: doesn't an ontological account make development unmeasurable, and therefore unaccountable? The opposite is true, and it is one of the model's practical advantages. Locating the gap between knowing and doing is what makes it measurable, because it tells you to stop measuring knowledge — which a leader can have in full while the gap stays wide open — and start measuring the coordinative behaviour that the inner shift is supposed to produce. A knowledge test cannot detect the gap; it is, structurally, blind to it. The layer-three markers can. An organisation that tracks how its leaders actually request, promise, and complete has a sharper instrument for the gap than any number of competency assessments, precisely because those markers move only when something below layer one has changed. Ontology, far from excusing development from accountability, is what finally lets development be held to it.
The knowing–doing gap is usually treated as a disappointment — the regrettable distance between good training and real behaviour. It is better treated as information. It is the field telling us, with great consistency, that the model under our development practice is wrong: that knowledge was never the thing that produced action, and that pouring more knowledge into the gap will not close it because the gap is not made of missing knowledge.
What produces action is who a leader is being, made effective through how they coordinate in language. Knowing, Being, Doing — three layers, governed by three different logics, and a gap that opens between the first and the rest whenever an organisation develops one and neglects the others. The leaders an organisation most needs are not the ones who know the most about leadership. They are the ones for whom effective action has become natural self-expression rather than a thing they remember to attempt. Closing the knowing–doing gap is the work of making more such leaders, on purpose, and being able to prove it. That is the work this desk exists to advance.
IMPACT THINKING RESEARCH · BY BEN BOTES · © 2026
If your leaders know exactly what to do and still don't reliably do it, this is the work. Tell us where it's showing up.
Start a conversation →