Organisation design conversations go in circles. Agile advocates argue about teams. Holacracy advocates argue about roles. Traditional designers argue about structure. The more sophisticated participants start talking about operating models, or leadership layers, or culture. They are all right — and all incomplete.
The problem is not a lack of good models. We have many: Galbraith's Star Model, Mintzberg's structural configurations, Laloux's Teal organisations, Holacracy's constitutional governance, the Viable System Model, and a dozen Agile scaling frameworks. The problem is that each model defines its own vocabulary. They are not translations of each other. They are not commensurable. A Holacracy practitioner and a traditional hierarchy designer are not disagreeing about which model is better — they are frequently speaking different languages about what an organisation is.
No shared grammar sits underneath them. No agreed set of primitives from which all these models are composed. When you move from a traditional hierarchy to a Holacratic structure, you do not know precisely what you are changing, what you are retaining, and what you are silently abandoning.
This article proposes that grammar. It consists of six atomic elements, one rule system, and two transversal conditions. Together, they describe the full coordination architecture of any organisation — not as a new model competing with the others, but as the vocabulary from which all models are built.
- The metaphor, and why it is not decorative
- Where this sits
- The architecture: Ownership, Grammar, Work
- The six elements
- Role
- Skill
- Group
- Decision
- Obligation
- Accountability
- Governance — the rule system
- Two universal tensions
- Culture is the emergent grammar
- The grammar in action — one comparison
- The grammar and the machine
- The grammar as OEF foundation
- Conclusion
- Comments and Feedbacks
- References
The metaphor, and why it is not decorative
A language grammar has parts of speech — nouns, verbs, prepositions — syntax rules for how they combine into legitimate sentences, and notation conventions for rendering those sentences on a page. The organisational grammar operates identically.
Elements are the parts of speech — the atomic design objects. Governance is the syntax — the rule system for how elements may be created, combined, modified, and dissolved. Organisational representation — org charts, network diagrams, process maps — is the notation. The org chart is not part of the grammar, just as the choice to write left-to-right is not part of English grammar. Confusing notation for grammar is one of the most persistent errors in organisation design. The org chart is a rendering; what it renders is the grammar.
Where this sits
The idea of a grammar of organisation is not new. The word has been used before, and in at least three distinct traditions that deserve acknowledgment.
The most famous use belongs to Karl Weick, who in The Social Psychology of Organizing (1979) defined organising as:
"a consensually validated grammar for reducing equivocality by means of sensible interlocked behaviors." Karl Weick
Weick's grammar is socio-cognitive — it describes how people assemble actions into sensible sequences, how they make meaning from ambiguity. It is a grammar of enactment. What I am proposing here is different in kind: a grammar of structure. Not how people make sense of organising, but what must be designed for coordination to exist. The two are complementary — Weick explains why organisational grammar is interpreted differently by every participant; the structural grammar explains what they are interpreting.
A second tradition is transactional. Gerald Salancik and Huseyin Leblebici, in work beginning in the late 1980s, proposed a genuinely generative grammar of organisational transactions — one that could algorithmically produce all possible organisational configurations from a finite set of rules governing resource dependencies and agent capabilities. Leblebici later extended this to constitutional rules of organising: rules of causal order, membership, allocation, and discourse. This work proved the concept of a formal generative grammar. Its primitives, however, are transactional — activities, resources, capabilities. The grammar I propose operates on a different substrate: the structural objects (Roles, Groups, Decisions) that transactional sequences flow through.
The third tradition is institutional. Sue Crawford and Elinor Ostrom's ADICO framework (1995) decomposes institutional statements into five syntactic components — Attributes, Deontics, Aims, Conditions, and Or-else clauses — producing a formal grammar that distinguishes strategies from norms from rules. ADICO maps remarkably well onto the Governance layer of this grammar: the three institutional types it identifies (shared strategies, norms, and rules) correspond to the three forms of Obligation defined below. ADICO governs the rule system; this grammar governs the structural objects the rules apply to.
One further intellectual debt is worth naming. John Zachman's enterprise architecture ontology (1987) introduced a distinction between primitives (timeless, single-variable elements) and composites (temporal, multi-variable implementations). That distinction is load-bearing here. Traditional organisation design operates almost entirely at the composite level — departments, job descriptions, reporting lines are all multi-variable bundles where roles, decisions, skills, and accountabilities are permanently fused. The grammar decomposes those composites into their constituent primitives. That is why it can describe radically different organisational forms using the same vocabulary: it works below the level at which models diverge.
Where does this leave the practitioner frameworks — Galbraith's Star Model, Mintzberg's structural configurations, the various consulting operating model toolkits? They are diagnostic instruments, not generative ones. They tell you what to assess, not what you are building from. The grammar does not compete with them. It sits underneath them, providing the atomic vocabulary from which their higher-level categories are composed.
The architecture: Ownership, Grammar, Work
The grammar does not sit in isolation. It operates between two transversal conditions: Ownership above, and Work below.
Work is the productive foundation — the totality of activity required to fulfil a purpose. Work is pre-grammatical. It exists before any organisational design is applied. It is the demand that makes the grammar necessary. The necessity principle runs as follows: when Work exceeds what a single individual can accomplish for an owner's purpose, coordination architecture is required. The grammar is that coordination architecture. The solo lawyer needs no grammar. The law firm with two hundred partners needs all of it.
Ownership is the authority frame within which all organisational design operates. Ownership determines who holds residual claim on the organisation's value and, through that claim, ultimate authority over its direction. It is not an element to be configured — it is the condition that determines which configurations are permissible. A cooperative, a private firm, a public institution, a platform business, and a family enterprise are five different ownership forms that produce five different feasible grammars from the same six elements.
I am indebted here to an insight from a workshop with Var Group in June 2026. In the middle of a discussion about self-management transformation, Francesco Frugiuele remarked simply: "the company still has an owner." That remark crystallised what I had been missing from much of the self-management literature. Self-management does not dissolve ownership — it operates within it. The first question of organisation design therefore is not "what structure?" but "whose organisation?"
The grammar fills the space between. When Work cannot be done by one person, and Ownership defines the frame, the grammar is what must be designed.
The six elements
Let me now define the six elements. I should be transparent about their origin: these six were not derived from first principles or deduced from a theoretical proof. They were arrived at through iterative testing — applying candidate element sets against real organisational configurations (hierarchies, self-managing systems, matrix structures, platform businesses, cooperatives) and asking a compositional adequacy question: can this set describe any organisation I encounter? Can it expose the differences between models that practitioners experience as real? Can I remove any element without losing explanatory power, or is there an organisational feature I cannot describe without adding one?
The claim is not that these six are the only possible primitives, or that they are logically necessary. The claim is compositional: together, they are sufficient to describe the coordination architecture of any organisation I have tested them against — and each carries weight that the others cannot absorb. If a more parsimonious set can do the same work, it should replace this one. But so far, none has.
Every organisation uses all six. The question is always how each is configured.
Role
Role is the smallest unit of expected contribution. A Role defines what a person — or, increasingly, a machine — is expected to do, know, and answer for, independent of who holds it and where it sits. Roles can be thick (bundling many responsibilities into a single package, as in traditional hierarchy) or thin (carrying a narrow, explicitly scoped set of accountabilities, as in Holacracy). How thick or thin Roles are designed to be is one of the most consequential grammatical choices available. When organisations talk about "job design", they are working in Role grammar — often without realising it.
Skill
Skill is the capability requirement attached to a Role or to Work. A Skill in the grammar is a design specification, not a personal attribute — it describes what capacity is needed, not whether a particular person possesses it. "This Role requires the ability to navigate executive stakeholders at Group level" is a grammatical statement. Whether a specific individual has that capacity is a staffing question that operates on top of the grammar. Skill is where the grammar connects to capability architecture — workforce planning, learning investment, and the increasingly consequential decisions about what to allocate to humans versus machines. Without Skill in the grammar, you can describe what needs doing but not what it takes to do it.
Group
Group is any assembly of Roles created to serve a collective purpose. This absorbs what we imprecisely call "teams", "units", "functions", and "departments" into a single element — one that varies along two dimensions. The first is organising principle: Groups can be assembled around shared work (product teams, project teams), shared skill (functions, chapters, communities of practice), financial planning (cost centres, profit centres), regulatory compliance (legal entities, statutory bodies), strategic positioning (business units, category organisations), or coordination across other Groups (regions, shared services, matrix overlays). The second dimension is structural permanence: from ad hoc collaboration to semi-recognised working group to formally constituted department with budget, mandate, and headcount authority. Most organisational confusion arises from not distinguishing why a Group was created. A cost centre asked to behave as a business unit is a grammatical error — and it produces the frustrations you would expect from a grammatical error.
Decision
Decision is the atomic unit of organisational action. A Decision is a commitment to a course of action that constrains future options. Decision placement — where in the organisation decisions are made — is arguably the single most powerful lever in organisation design. Push decisions upward and you create bottleneck and learned helplessness. Distribute them to the point of Work and you gain speed but introduce coordination risk. The grammar does not prescribe which configuration is right. It makes the trade-off explicit, and it makes it comparable across models.
Obligation
Obligation is the connective bond between elements, carrying mutual expectations. Obligations run between Roles, between Groups, between the organisation and its ecosystem partners, and between the organisation and individuals. They take three forms: contractual (formalised with legal standing), normative (socially enforced without legal standing), and implicit (assumed but unstated — the psychological contract, the unwritten expectation about availability or loyalty). The direction of Obligation is one of the most revealing features of any organisation. Hierarchies create primarily upward obligations. Self-managing systems create lateral ones. Obligations also run beyond the organisation's boundary — and their direction there is equally revealing. B2B organisations carry obligations directly to end customers. B2C organisations carry them through trade intermediaries. Where the Obligation grammar is absent or ambiguous, conflict reliably fills the gap.
Accountability
Accountability is the assignment of answering for an outcome. Accountability attaches to a single element — this Role, this Group answers for this result. It differs from Obligation in directionality: Obligation runs between elements as a mutual bond; Accountability is assigned to an element as a burden of consequence. The most diagnostically useful property of Accountability is its relationship to Decision. Being accountable for an outcome you cannot fully decide is one of the most common sources of organisational frustration — and one of the most precise things the grammar can expose. The brand manager accountable for brand performance but unable to decide trade spend. The middle manager accountable for strategy execution but unable to decide resource allocation. The human operator accountable for an AI system's output but unable to understand or override the algorithm. Every one of these is a grammatical sentence with a Decision-Accountability misalignment at its core.
Governance — the rule system
The rule system that determines how elements may be created, combined, modified, and dissolved — and who has the right to do so — is Governance. It is not an element. It is the syntax of the grammar.
Different organisational forms produce radically different governance structures from the same six elements. Hierarchy governs through escalation: decisions travel upward until they find sufficient authority. Holacracy governs through constitutional process: Roles and Circles can only be changed through a defined governance meeting. Agile governs through cadence: stand-ups, sprint reviews, and retrospectives create regular decision points. Cooperatives govern through democratic process. In each case, the elements are the same. The syntax governing their combination differs entirely.
Authority — often treated as a separate concept — lives inside the Governance layer. Authority is a governance assignment: the right to make certain Decisions, create certain Roles, or dissolve certain Groups. When organisations struggle with empowerment, what they are usually discovering is that the Governance layer has not been redesigned to match the stated intent. The language says "distributed authority"; the grammar says "escalation hierarchy". The grammar wins.
Two universal tensions
Every element is subject to two tensions that provide diagnostic power wherever you apply them.
Let me say them plainly. The first is formalisation — the spectrum from informal practice to institutionalised structure. Roles can be formally designed or informally assumed. Groups can be ad hoc collaborations or chartered departments — which is the structural permanence dimension described above, now seen as an instance of the general tension. Obligations can be handshakes or contracts. Where each element sits on this spectrum is a design choice; or, if it is not a design choice, it is an accumulation. The pattern of formalisation across all six elements characterises an organisation's operating style more accurately than any single structural label.
The second is intentionality — designed versus emergent. Every element can be deliberately configured or can arise without anyone deciding. Roles get designed, but shadow roles emerge. Groups get chartered, but informal networks form around trust. Obligations get contracted, but implicit expectations accumulate. Decisions get governed, but corridor decisions happen constantly.
The gap between the designed grammar and the emergent grammar is where organisational reality actually lives. No organisation operates purely on the grammar it designed. Every one accumulates an unofficial grammar alongside the official one. The diagnostic power of formalisation and intentionality, applied across all six elements simultaneously, is that they reveal the shape of that gap — and the shape of that gap tells you more about the organisation than either the formal design or the informal reality alone.
Culture is the emergent grammar
Culture is not a separate phenomenon from the grammar. It is its emergent face — what the OEF calls Corporate Culture is, structurally, the set of informal roles, implicit obligations, shadow decisions, and unwritten accountabilities that operate where formal design stops or is not enforced.
When someone says "the culture here is hierarchical," what they mean, in grammatical terms, is that Decisions flow upward regardless of formal placement, Obligations run toward seniority, and Accountability is enforced through escalation. Culture-talk is grammar-talk without the precision. And the precision matters, because it changes what you do about it. Culture change is grammar redesign: reconfiguring the formal elements so that the emergent behaviour the organisation actually wants becomes the path of least resistance.
Language is the evidence. Listen to how people describe getting things done. "Let me check with my boss" reveals an escalation grammar. "That's within my circle's domain" reveals a constitutional one. "We'll take it to the retro" reveals a cadence-based one. The vocabulary people reach for when they describe how decisions are made, who they owe things to, and what happens when something goes wrong — that is the audit trail of the grammar actually in use. This is, in effect, a form of Organisational Awareness — the capability to perceive how the organisation actually operates, rather than how it imagines itself to operate.
The grammar in action — one comparison
Let me show the grammar at work on one comparison. Three organisation models — Traditional Hierarchy, Holacracy, and Agile — drawn from the same six-element vocabulary.
- Traditional Hierarchy makes Group (the organisational unit) and thick Role primary. Skill and Decision placement are largely invisible — assumed to follow from role title and position in the hierarchy. Accountability is individually assigned but often attached to Roles several steps removed from where Decisions are actually made. The Decision-Accountability gap is structurally embedded, not an aberration. Governance is escalation: legitimacy flows upward.
- Holacracy makes Role primary and decomposes to maximum thinness. Every Role carries an explicit purpose, a domain, and a set of accountabilities — the most granular Role grammar in mainstream use. Decision rights are distributed to the Role, not the person. Governance is constitutional: Roles and Groups (called Circles) can only be changed through a defined process. Holacracy is nearly silent on Skill — it specifies what accountabilities a Role carries but does not address what capabilities are required to fill it effectively.
- Agile (Scrum) makes Group (the cross-functional team) primary and elevates Work visibility through the backlog. Accountability is collective within the team boundary. Governance is cadence: stand-ups, reviews, and retrospectives are the decision rhythm. Agile — at least as practised in single-team Scrum — has the richest process grammar of the three, but it is a local grammar. It operates within the team boundary and largely leaves the inter-team coordination problem underspecified. Scaling frameworks (SAFe, LeSS, Nexus) attempt to fill this gap, but their track record is uneven precisely because they are layering coordination mechanisms on top of a grammar that was designed for team-level autonomy. The elements that must be designed across team boundaries — Obligation, Decision, and Group — require explicit grammar that Scrum itself does not provide.
What this comparison reveals is something conventional model descriptions obscure: all three models use all six elements. What differs is which elements are foregrounded, how granularly each is specified, and what governance applies. Calling Holacracy "flat" or Agile "autonomous" are claims about specific elements in specific configurations — not claims about the presence or absence of structure. The grammar is always there. The question is always whether it has been designed or merely accumulated.
The grammar and the machine
One question the grammar must address, because it is the question that most organisation design conversations are circling right now: what does AI actually change?
The conventional framing — that AI "transforms work" or "augments human capability" — is too imprecise to design from. The grammar makes the question answerable by decomposing it into specific elements.
AI reconfigures Role. When a machine can perform tasks that were bundled into a human Role, the Role does not simply shrink — it changes character. The analyst who previously spent time gathering and structuring data now spends time reviewing and interpreting AI-generated output. The Role has shifted from execution to oversight, and the Skill profile has shifted with it. Whether that shift is acknowledged in the formal grammar or left to accumulate informally is a design choice with significant consequences. In most organisations today, it is accumulating.
AI fragments Decision. Algorithmic systems make consequential choices — what to recommend, what to flag, what to filter — at speeds and volumes that no human governance structure can match. Decision rights that were previously attached to Roles are being absorbed by systems, often without any formal reassignment in the governance layer. The grammar still says the human decides. The reality is that the machine decides and the human ratifies, or the machine decides and no one reviews at all. This is not a technology problem. It is a grammatical incoherence — a mismatch between where Decisions are formally placed and where they actually occur.
And that incoherence produces the most dangerous misalignment the grammar can expose: Accountability without Decision. I described one instance earlier — the human operator accountable for an AI system's output but unable to understand or override the algorithm. It is not an edge case. It is becoming a structural feature of AI-augmented organisations. The brand manager accountable for campaign performance where the algorithm selects the audience. The hiring manager accountable for diversity outcomes where the screening tool shapes the shortlist. In each case, Accountability sits with a human Role while the Decision has migrated to a machine — and the Governance layer has not been redesigned to close the gap.
The grammar does not resolve these tensions. What it does is name them precisely enough to design around them. An organisation that is serious about AI adoption needs to audit three things: which Decisions have moved from human Roles to algorithmic systems, which Accountability assignments still assume those Decisions are human, and whether the Governance layer specifies how machine Decisions are monitored, overridden, and accounted for. That audit is grammar work.
The grammar as OEF foundation
The grammar is not a new addition to the Organisation Evolution Framework. It is its foundation.
The OEF's eight components are not eight separate things. They are eight analytical lenses on the same underlying grammar, each foregrounding different elements. Strategy foregrounds Decision and Group. The Operating Model foregrounds Role, Skill, Obligation, and Decision. Leadership foregrounds Accountability and Decision. The elements are the same; the lens determines which are primary.
When two OEF components conflict — when the operating model demands one configuration and the organisation model demands another — the grammar locates the conflict precisely. It is always a conflict about how a specific element should be configured. Naming it precisely is the first step toward resolving it.
OEF Component | Primary Elements | The lens asks |
Full grammar | How to configure for value creation and capture? | |
Decision, Group, Skill | Does the grammar support where we compete? | |
Role, Skill, Obligation, Decision | How does work flow, and where do handoffs fail? | |
Role, Group, Accountability, Decision | How are roles assembled and authority distributed? | |
Accountability, Decision, Role, Skill | How is direction set and accountability carried? | |
Obligation, Accountability, Governance | Does purpose create real constraints on design? | |
Emergent face of all six elements | What operates where formal design stops? | |
Obligation, Group, Governance | How do separate grammars interface at boundaries? |
This article is part of a series where I examine the different components of Organization Design, analyse models and theories and propose a specific approach of Intentional Design.
Key Content Articles
- Organisational Grammar - The Atomic Elements of Designed Coordination
- The Organization Evolution Framework
- Business Models: The Theory and the Practice
- Strategy. Frameworks: The Theory and the Practice
- Operating Models: The Theory and the Practice
- Organization Models: a Reasoned List between Old and New
- Leadership Models: The Theory and The Practice
- Purpose: The Theory and the Practice
- Corporate Culture: The Theory and the Practice
- Organization Ecosystem: The Theory and the Practice
- Building the Intentional Organization
- What is Organization Design?
- Consistency and Intentional Design: Building the Organization of the Future
Conclusion
The grammar does not tell you which organisation to build. That depends on your business model, your strategy, your ownership form, and the nature of the work to be done. What it does is clarify what you are actually deciding when you design. It gives you a vocabulary precise enough to compare models, expose trade-offs, and locate the source of dysfunction when it appears.
Next time you find yourself in an organisation design conversation that feels circular, try two questions.
- First: which elements are we configuring?
- Second: what rules govern how they combine?
If you can answer both, you are designing. If you cannot, you are using the language of organisation design without its grammar — and the conversation will keep going in circles.
Comments and Feedbacks
References
Read More