The Learning OrganisationWhat Senge's Idea Became

In the third chapter of The Fifth Discipline, the 1990 book that launched the idea of the learning organisation, Peter Senge describes a game first developed at MIT’s Sloan School of Management in the 1960s. Players run a simple beer supply chain, from retailer through wholesaler to brewery. Nobody may speak to anyone else, and each player sees only their own orders and stock. A single, modest change in customer demand is introduced. Within a few simulated weeks the chain is in chaos. Backlogs build and orders multiply in panic, until the game ends with warehouses full of beer nobody wants.

The game has been played thousands of times with very different groups of people, and it produces the same pattern almost whoever plays. Afterwards the players blame each other. The retailer is sure the wholesaler was incompetent; the brewery assumes someone downstream was hoarding. Senge’s point was that the behaviour came from the structure, and that people inside a structure find it very hard to see. What they see is each other.

The Fifth Discipline: The Art and Practice of the Learning Organization made the learning organisation one of the defining management ideas of the 1990s. Thirty-six years on, the phrase is everywhere, in values statements and the names of whole departments, while the thing it described remains hard to find. The distance between the two says a good deal about what organisations do with ideas that ask them to see themselves.

What Senge Meant by the Learning Organisation

Senge built the idea on five disciplines. Personal mastery, mental models, shared vision and team learning were held together by the fifth, systems thinking: the capacity to see patterns and feedback loops where we habitually see isolated events and people to blame. A learning organisation, in his account, was one in which people kept expanding their capacity to create what they cared about, and in which “collective aspiration is set free”.

The kind of learning he had in mind went beyond taking in information. Senge borrowed the Greek word metanoia, a shift of mind, for it: the experience of seeing a familiar situation differently and being unable to go back to the old way of seeing it. That is an unsettling thing to invite into an organisation. A shift of mind in the right person can make a strategy look foolish, or reveal that a successful department has been succeeding for reasons nobody wants to examine. Senge knew this, and much of the book is about why organisations defend themselves against exactly the learning they say they want.

Read now, what stands out is how much he was asking of the people at the top. The discipline of mental models meant surfacing the assumptions that shape what an organisation notices, including the assumptions held by those with the most authority to protect them. For team learning Senge drew on the physicist David Bohm’s work on dialogue, and Bohm’s conditions were demanding. Participants had to suspend their assumptions, holding them up to be examined instead of defending them, and they had to regard one another as colleagues. Senge added that everyone involved had to value dialogue more than their privileges of rank. Leaders, in this picture, designed the conditions in which others could learn and helped people see reality more clearly. They were not the people who already knew.

That last point carries the idea’s hardest demand. A learning organisation has to be able to not know things in public. Learning begins with the admission that your current picture is incomplete, and most organisations are built to reward the opposite: the confident plan and the leader who has the answer. Senge was asking organisations to treat unknowing as a normal condition of working life, something to be held and explored, where most treat it as a weakness to be hidden before the next board meeting. It was an encouraging book with a hard request inside it.

How the Learning Organisation Became a Programme

Ideas that ask organisations to change how they see are usually converted into things organisations can do. In 1993 David Garvin published “Building a Learning Organization” in the Harvard Business Review. He argued that writing on the subject had been “reverential and utopian”, heavy on near-mystical language and light on guidance a manager could use. His alternative was practical: a set of activities, from systematic problem solving to moving knowledge around the company, and an insistence that learning had to be measured.

Garvin was right that managers needed more than inspiration, and the article is still widely read for good reason. It also marks the moment the idea turned. Once the learning organisation had to be measured, it had to become something that could be counted, and what can be counted is mostly activity. By the late 1990s there were questionnaires to score how much of a learning organisation you were; in 2008 Garvin, Amy Edmondson and Francesca Gino published another in the same journal. Meanwhile learning moved into a department, with its own systems and completion rates, and a line about a learning culture in the values statement. Each of these has its uses. Taken together, they turned an idea about an organisation examining its own thinking into a function that supplies content to individuals.

Gregory Bateson’s levels of learning help explain what was lost. At the level Bateson called Learning I, people learn content: this response produces this result. At Learning II, which he also called deutero-learning, they learn the context: what kind of situation this is, what gets rewarded here, what is safe to say. Everyone in an organisation is engaged in Learning II all the time. A new starter learns within weeks which questions are welcome and which end careers, and whose ideas tend to be credited to someone more senior.

The learning organisation, as implemented, addressed Learning I. It offered people more content. Learning II carried on underneath, teaching people what the organisation valued in practice, and when the two disagreed Learning II won, as it generally does. A company can run a hundred courses on innovation and still teach, through who gets promoted, that the safest move is to repeat what worked last time.

Chris Argyris had a name for this gap: the difference between espoused theory, what an organisation says it believes, and theory-in-use, the beliefs its behaviour reveals. Applied to learning itself, the espoused theory is that the organisation values learning. The theory-in-use is that errors are costly and uncertainty is best kept out of sight. The mythology of learning from failure is one of the places the gap shows most clearly, in the review meeting whose lessons were settled before anyone sat down.

The language changed along with the practice. In Senge’s usage learning was continuous and often uncomfortable, something that happened to the people doing it and left them seeing differently. In most organisations the word now refers to things that are finished. There are lessons learned, captured in a document at the end of a project, and there are “learnings”, a noun nobody needed, presented on a slide. Both treat learning as a product, complete and ready to hand over, which is the form an organisation can absorb without changing anything else. Words drift this way when an idea is taken up by the system it was meant to question, and the meaning this one now carries has moved a long way toward compliance.

The newest version of the idea runs on AI. Organisations increasingly describe AI as the way to capture institutional knowledge: record what experienced people know, make it searchable, stop it walking out of the door when they leave. It is the learning organisation’s oldest ambition in its most literal form, organisational learning without the inconvenience of the people who do it, and it inherits the old blind spot. What can be captured is what can be written down, which is Learning I. What the organisation has learned about itself, including whose judgement really counts, stays where it always was, in people and in the relationships between them. That is also where most of what AI is doing to organisations is currently happening.

Why Power Limits the Learning Organisation

The deepest difficulty was always power, and it was visible early. In the mid-1990s the management scholar John Coopey pointed out that advocates of dialogue tended to overlook how unevenly people are able to set the terms of a debate. Dialogue among colleagues is one thing. Dialogue between a manager and someone whose performance review that manager will write next month is something else, however skilled the facilitator. Bohm’s condition that participants regard one another as colleagues describes a relationship, and hierarchy is a standing arrangement that makes the relationship hard to sustain.

Shared vision raises the same question in another form: whose vision? Senge was clear that a genuinely shared vision could not be imposed from above. In practice many organisations developed a vision at the top and then measured how far it had been taken up below. People were asked to hold as their own a picture they had little hand in forming. Most did the sensible thing and learned to talk about it fluently, which is Learning II doing exactly what it does.

Power also decides who can afford to be seen not knowing. A senior leader who says “I don’t know” in a meeting is often admired for openness. A junior colleague who says the same thing is more likely to be marked down as not yet ready. Amy Edmondson’s early research in hospitals found that the better-led teams appeared to make more medication errors, and the reason turned out to be that their members felt safe enough to report them. Learning depends on that kind of safety, and safety is distributed along the same lines as authority. Psychological safety has since travelled the same road as the learning organisation, from a finding about how groups work to a programme with its own survey, which suggests the pattern lies in what organisations do with ideas of this kind.

Ralph Stacey’s critique goes further. Senge’s systems thinking asks the leader to step back far enough to see the whole system, and then to redesign it. Stacey argued that nobody gets to step back. Leaders are participants in the patterns they want to change, caught up in the same anxieties and the same local conversations as everyone else, and what emerges from those conversations can’t be designed in advance. On this view learning happens between people, in relation, and “the organisation learns” is a shorthand that hides who did the learning and who decided which parts of it counted. Read through Stacey, the learning organisation kept the machine model it was meant to replace and put a more thoughtful engineer in charge. Thinking of organisations as complex adaptive systems starts somewhere else entirely, with no engineer at all.

There is also a belonging question folded into the word “learner”. In a 2001 paper titled “Against Learning”, the organisation theorist Chris Grey argued that learning talk tends to individualise: it makes each person responsible for their own adaptability. Inside organisations this becomes a condition of membership. The good employee embraces change and is always learning. Someone who questions a change can be recast as someone who has failed to learn, and their objection can then be handled as a development need. Learning, which Senge meant as a shared capacity, became a personal virtue that could be used against the people it was supposed to serve.

The same move now runs through talk of reskilling for AI. When roles change, the responsibility for keeping up falls on the individual, and a person who cannot reinvent themselves on the organisation’s timetable has, by this logic, chosen to be left behind. The decisions that set the timetable, and the people who took them, rarely appear in that account.

What Senge Got Right About Learning Disabilities in Organisations

Senge’s diagnosis has lasted better than his cure. Alongside the five disciplines he described seven learning disabilities, and several read today as plain description of organisational life. “I am my position” names the way people come to identify with their role so completely that they stop seeing how it connects to anything beyond it. The delusion of learning from experience names the fact that the consequences of our most important decisions often unfold far away in time, or elsewhere in the system, where we never experience them directly. The myth of the management team names the senior group that presents a united front, avoids the disagreements that would expose its uncertainty, and becomes very good at what Argyris called skilled incompetence.

None of these requires anyone to behave badly. That was always Senge’s most useful insight, and it is the one the beer game makes vivid. Ordinary people doing their jobs sensibly, inside a structure none of them can see whole, produce outcomes none of them intended and then blame each other for. The learning organisation was meant to be the answer to that. In most places it became another part of the structure, a department with a questionnaire attached, sensible in every detail and one more thing people work inside without seeing.

Senge even supplied the terms for what happened to his own idea. One of the recurring patterns he called systems archetypes is shifting the burden. A problem appears, and a quick fix relieves its symptoms. The relief takes away the pressure to deal with the underlying cause, and over time the organisation’s capacity to deal with that cause weakens, until the quick fix is all it has. The learning programme fits the pattern closely. It relieves the symptom, the uneasy sense that the organisation ought to be learning, and in doing so it removes the pressure for the harder change: an organisation able to examine how it thinks, and whose thinking it has decided to count.

When the beer game ends, the facilitator shows the players the structure that produced their behaviour, and most of them feel relief. It wasn’t their fault. Senge hoped the relief would turn into curiosity about the systems people build and live inside. The question it rarely turns into is who set the game up. In the classroom the answer is simple: the facilitator did, and the rules are printed on a sheet. In an organisation there is no sheet. The rules were written by many hands over many years, some of them belonging to the people now playing, and nobody stands outside to hand them round.

A learning organisation would be one that could ask that question about itself and stay with the answer long enough to be changed by it. Thirty-six years on, it is still not clear what that would look like, or who would be allowed to ask.