Task differentiation

In the most basic form of division of labour, a practical undertaking is decomposed into contributory activities and those activities are assigned directly to individual participants. Call this task differentiation. The differentiation terminates in assignment of a portion of the labour itself.

Task differentiation requires that the practical undertaking admit of decomposition, participants capable of performing component tasks or activities, and some mode of coordination among them – the last can be as minimal as temporal sequencing or as demanding as real-time mutual adjustment. It does not require that participants occupy positions of relative authority over one another, or that the arrangement persist beyond the period in which tasks are performed.

The diagnostic mark of task differentiation is therefore that responsibilities attach to participants in the first instance, not to person-shaped functions or positions that outlast the undertaking. What is assigned is the labour contribution itself.

The emblematic case is the bucket brigade: strangers at a fire form a line for passing buckets, each taking up a place in a chain that lasts only as long as the emergency. The particular brigade disperses when the fire is extinguished. The other familiar instances are variants on it. An ad hoc group pushes a car out of a ditch; hikers divide who carries the water and who reads the map; and, from the literature on human evolution, Ofek’s (Reference Ofek2001) fire-keeper takes on the dedicated function of maintaining the cooking fire in a proto-human village while others hunt or forage. Each serves an immediate practical need and attaches responsibility to a participant rather than a standing position.

The particular assignments lapse when the purpose is served or the group disperses, but the pattern they establish can remain available as precedent. If a second fire starts nearby, attendees at one fire may reassemble the brigade by repeating a pattern learned in the first episode without thereby constituting roles between fires.

Role differentiation

While task differentiation assigns contributed primary labour directly to participants, many arrangements instead organize labour by assigning participants to person-shaped positions. Call such a position a role, and the resulting form role differentiation. A role supplies a standing basis for allocating further tasks, permissions, responsibilities, jurisdictions, and authority as circumstances arise.

Occupancy of a role is itself governed by conditions of eligibility. These may include competence or qualification, as in many professions, but need not: appointment, election, seniority, heredity, or other socially recognized conditions can likewise determine who may occupy the position. When a hospital distinguishes nurses, residents, and attending physicians, the relevant unit of assignment is not a task as such but a role to which tasks might subsequently redound. The position exists without regard for its current occupant and carries duties, permissions, jurisdictions, responsibilities, and authorities that bear on where further duties, prerogatives, and leadership will fall.

Because the role is not identical with its incumbent, a role-differentiated structure can persist beyond any particular episode of need or turnover among occupants: when an incumbent departs, the position may remain as a vacancy to be filled, hence the need for mechanisms by which positions are filled and (where necessary) vacated – which may be formal and explicit, as in a bureaucratic organization, or informal and tacit, as in the (currently defunct) guild system.

Replaceability of incumbents is therefore a characteristic consequence of role differentiation. The role is person-shaped but not person-specific: the arrangement need not dissolve when a participant departs.

Role differentiation comprises at least two sub-forms that deserve mention. In organizational role differentiation, positions are defined by an explicit authority structure – an ordered arrangement of offices with specified duties, jurisdictions, and lines of accountability. The paradigm case is Weber’s bureaucracy, but the form extends to any standing organization in which roles are formally constituted and officially assigned: the hospital, the law firm, the military unit, and the university department. In professional role differentiation, positions are defined not by an organization’s internal authority structure but by a community of practice that controls entry, certifies competence, and polices the boundaries of the role’s legitimate jurisdiction. The paradigm case is Abbott’s (Reference Abbott1988) profession, where the relevant unit is not the office but the licensed expert whose authority derives from demonstrated mastery of a recognized domain rather than from appointment within an authority structure.

These two sub-forms share the defining structure of role differentiation – person-shaped positions that supply a standing basis for allocating responsibilities and authority – while differing in how the position is constituted, who is eligible to occupy it, and how it is reproduced, as we will discuss below.

Why role and task are both taxonomically basic

Differentiation by role is no subspecies of differentiation by task. A task assignment allocates a contribution to be performed, whereas role assignment establishes a standing basis for allocating further tasks, permissions, responsibilities, and authority as circumstances arise. Once roles are in place, the question of allocating tasks is no longer answered from scratch: the existing role structure bears on where newly arising demands properly land. A role may therefore encompass tasks that could not have been specified when the role itself was first constituted – it is now part of the logic by which subsequent labour and authority are distributed.

Two features follow.

• • Role norms and responsibilities. Roles carry duties, permissions, jurisdictions, and relations of accountability that govern the allocation of particular tasks without being exhausted by them. The attending physician’s standing, for example, determines which decisions and responsibilities fall to that role as cases arise, including circumstances not anticipated in advance.

• • Anticipatory structure. Because roles carry within them rationales for assignment of not-yet-allocated and even not-yet-conceptualized duties, permissions and leadership, they allow an arrangement to defer allocations of duties, prerogatives and permissions, even ones that can be envisioned now.

None of this implies a natural progression from task differentiation to role differentiation. The two are distinct organizational forms. A task-differentiated arrangement may remain task-differentiated; role differentiation arises only where a standing structure of assignment is established or considered desirable. And both forms of division can coexist simultaneously, governing divisions of different undertakings by the same participants.

Station differentiation

Task and role differentiation both organize contributions to undertakings or standing functions. A third form concerns not an episodic contribution or a recurring one, but a participant’s place in a broader order governing access to work, burdens of work, permissions, exemptions, and claims on its surplus. Call this station differentiation. A station need not be a function in relation to an acknowledged enterprise of the group: its significance lies in the relative social placement it confers. Caste, estate, classed access to occupations, and gendered allocations of paid and unpaid labour supply familiar cases. Someone might resist labelling station differentiation as a form of division of labour, preferring to treat stratification as overlaid upon division of labour schemes. But stratification determines who is eligible for a role or task, who is exempt from particular roles or tasks, who is expected to undertake or withhold undesirable or ‘support’ labour, and who is entitled to the largest share of the resulting surplus of the division scheme. In that strict sense it bears quite fundamentally on division of labour.

Importantly, station differentiation is not role differentiation writ large. Roles are individuated by functions in a standing structure; stations by priorities assigned to participants in an order of access, burden, permission, and claim. A role presents a person-shaped vacancy that different incumbents may fill. A station does not present any vacancy, but a relationship of eligibility for or vulnerability to one or more: one can be related to a role or task as landed or landless, male or female, black or white, adult or dependent, without thereby occupying any office, possessing any qualifications, or aspiring to any relevant ambitions.

The mechanisms by which stations are maintained or reproduced are downstream questions about integration and persistence rather than further features of station differentiation. Social closure and social reproduction will therefore reappear in the section Selection, persistence, and change, as processes bearing on the durability of particular station arrangements.

Three placements of coordinative labour

In non-mediated integration, those performing the differentiated labour also perform the work of making it cohere. Two people carrying a table through a doorway continually accommodate one another; a string quartet gives way and presses forward through reciprocal attention. The parties to the enterprise themselves bear the coordinative burden. Such coordination can be demanding in time, attention, and social skill, and its costs rise as the number of participants, distance, or temporal separation between them grows.

In intermediated integration, coordinative labour has itself been differentiated and assigned. A group may orient around a leader; a foreman may direct a crew; an organization may construct offices and procedures whose work is to coordinate the activities of others. The intermediary is identifiable even where, as in a bureaucracy, the coordinating entity contains a substantial division of labour of its own.

In diffuse integration, by contrast, no worker-coordinator performs the integrative work. Coordination is performed instead by many proxies: prices and other signals, professional or social expectations, customary classifications, received wisdom or convention, and sedimented patterns of conduct. Such coordination may be highly effective without a single agent performing integrative work.

Thomas Schelling’s famous (Reference Schelling1960) discussion of focal points makes one aspect of this phenomenon vivid. Participants may converge without direction or explicit agreement because social expectations raise the cognitive availability – the salience – of environmental or social cues, as each anticipates the anticipations of the others. This is a process of reverberating anticipation, in which participants take the measure not simply of what others think or express the intention to do, but of what others take to be shared as guidance in coordination. For present purposes, we thus draw attention here to coordinative work that may be performed by the surrounding field of signals, expectations, and socially available saliences: they can constrain and align conduct without themselves enjoying agentic powers.

The three placements can coexist at different scales in a single activity. A hospital ward may combine mutual adjustment among clinicians, administrative procedures, and the cadre of professional norms that is the preserve of no one single hospital.

Placement alone does not identify mechanism. We turn now to mechanisms within each family.

Mechanisms within the three placements

The first non-mediated mechanism is mutual adjustment: the parties coordinate in the course of performance by orienting and responding directly to one another. The bucket brigade and string quartet are paradigms. The philosophy of joint action addresses how divided activity is coordinated by the participants themselves – through meshing plans, joint commitment, or related structures (Bratman, Reference Bratman1987, Reference Bratman2014; Gilbert, Reference Gilbert1989, Reference Gilbert2014; Ludwig, Reference Ludwig2016; Roth, Reference Roth2017; Tuomela, Reference Tuomela2007). Not every joint action divides labour, but where it does, these accounts describe one important mechanism of non-mediated integration.

A second non-mediated mechanism is negotiated coordination. Here the parties themselves explicitly settle an allocation, plan, sequence, or set of terms that organizes their later contributions. Hikers may decide who carries water and who safeguards the map; collaborators on a research project may divide up responsibilities before beginning work. Unlike mutual adjustment, the principal coordinative work of negotiated coordination need not be performed continuously in real time. The mechanism is non-mediated because it is the parties themselves who settle the terms rather than an entity tasked with coordination.

Our first intermediated mechanism is direction, in which coordinative labour is assigned to a coordinator who exercises discretionary judgement about what those contributing first-order labour are to do. Direction includes emergent leadership, such as when a group looks to one of its number for direction, perhaps on the basis of experience, charisma, or trust. Command is a limiting form of the same mechanism, in which direction is backed by a standing office or preexisting authority (DeRue and Ashford, Reference DeRue and Ashford2010; Ganz, Reference Ganz2009; Hollander, Reference Hollander1992; Weber, Reference Weber1922). Marx’s workshop case and Simon’s employment relation isolate this placement from different angles: the former by contrasting organization via command with market coordination, the latter through the employee’s acceptance of direction within what Simon calls an “area of acceptance”. The wildcat strike that throws up a spokesperson and the work crew directed by a foreman differ in the grounding of direction while sharing its basic mechanism.

The second intermediated mechanism is administered rule. Here a standing organization coordinates through offices, procedures, schedules, jurisdictions, and rules that organize conduct without requiring case-by-case direction. Bureaucracies, civil services, hospitals, universities, and firms are familiar cases of administered rule (Miller, Reference Miller2019; Weber, Reference Weber1922). Weber’s bureaucratic office, introduced in the section Competing accounts and their limits as a paradigm of role differentiation, reappears here from the second axis: the office is a person-shaped unit of assignment, whereas the bureaucracy’s offices and procedures provide an intermediated mechanism of integration.

We proceed now to mechanisms of diffusely coordinated labour. The first diffuse mechanism is signal-mediated coupling. Signals generated across a wider field register dispersed activity and become inputs to further activity, thus reverberating: participants adjust to them, and those adjustments in turn alter the signals confronting others. Coordination is orchestrated by this distributed feedback rather than by any explicit direction or plan on the part of any agent. Hayek’s (1945) price system is the paradigm: a new use for tin raises its price, buyers automatically economize, and their responses propagate as further information. Exchanges, market-makers, contract law, property arrangements, and platforms help constitute the institutional setting in which actual market signals are generated and interpreted; within that setting, the price mechanism coordinates through decentralized responses rather than at the direction of entrants.

The labour market makes the cross-cutting character especially clear. A discrete task – a paint job, editing a manuscript, harvesting a field – may be allocated through spot contracting. The unit of differentiation remains a task, whereas its allocation is integrated diffusely through offered wages, bids, availability, and other market signals. The same task may be folded into the operations of a single firm, where it will be subject to a foreman or schedule, and the integration becomes intermediated without necessarily moving on the primary axis.

The firm/market boundary sketched in the section Competing accounts and their limits can now be stated more exactly. Coase (Reference Coase1937) raises the question of why some transactions are organized within firms rather than through markets; transaction-cost analysis explains selection of mechanism by comparing the costs of securing coordination through market exchange with those of internal organization under relevant conditions. The present taxonomy identifies the alternatives between which selection occurs. Market contracting leaves the task’s integration diffuse, orchestrated by prices, bids, and other signals; bringing the same task within a firm places much of its coordinative labour in an organization, typically through direction or administered rule. Placement therefore describes the structural change to be explained; it does not compete with an economic account of why one arrangement is advantageous under specified conditions.

The second diffuse mechanism is sedimented normative coordination. Participants encounter a field already patterned by norms, conventions, classifications, expectations, and precedents that make some assignments, responses, and claims socially ordinary and others anomalous. Precedent belongs here because what participants have done before can itself supply a reason to repeat an allocation or sequence – because it worked, because it has become familiar or settled, and sometimes even when the earlier arrangement worked poorly. Here coordination is orchestrated diffusely through what participants take to be customary, fitting, expected, or settled, rather than through continuously updated feedback from current activity or case-by-case direction from a coordinator.

Sugden’s (Reference Sugden1989) account of spontaneous order supplies a complementary economic case: rules can evolve and sustain coordination without deliberate design or formal machinery of enforcement. This differs from the self-governing arrangements considered below (see the section Vigilance and revising labour), in which participants retain practical capacity to monitor, contest, and revise the rules that govern them. In the present terms, spontaneous order is not a fourth placement but one way diffuse coordination via norms can arise and persist.

The field-level templates emphasized by DiMaggio and Powell (Reference DiMaggio and Powell1983) and the ready-made category distinctions that Tilly (Reference Tilly1998) sees organizations importing are examples of such already available coordinative materials. Caste-linked occupational assignments, inherited expectations about gendered care labour, professional custom, and other customary allocations are already socially available when participants encounter them; their conduct may reproduce or alter the pattern, but does not have to generate them from scratch.

The bucket brigade example supplies a minimal case. The first brigade may disband, yet its pattern can serve as precedent for reassembling a later brigade at a second fire without any role persisting between episodes. For task differentiation, an unspoken convention can allocate a one-off contribution – for example, that whoever cooks is spared cleanup – without negotiation or direction.

When it comes to roles, however, professional practice or other custom can organize how physicians, nurses, or lawyers hand work across standing jurisdictions. And where station is concerned, cultural expectations can align gendered care labour or caste-linked occupational access.

The rules/equilibria dispute can now be revisited from a different angle. Hindriks and Guala (Reference Hindriks and Guala2015) argue that equilibrium-based and rule-based accounts are individually inadequate and propose a unified conception of institutions as rules-in-equilibrium. Hodgson (Reference Hodgson2015) accepts the explanatory importance of equilibrium analysis but argues that equilibria are behavioural outcomes and should not enter the taxonomic definition of institutions, which he defines in terms of rules.

The present taxonomy is silent on the matter in dispute here. A rule does not, as such, determine where or how coordinative labour is performed: rules may be administered through offices or operate diffusely as norms, conventions, or precedents. Nor does an equilibrium pattern determine placement: settled patterns of conduct can be sustained by a variety of distributions of coordinative labour. Placement therefore cross-cuts the terms of the controversy. The present taxonomy thus leaves the disagreement between the parties, as to the definition of institution intact while identifying a structural question neither answer settles. Whether an institution is essentially a rule in equilibrium, a system of rules whose operation may be illuminated by equilibrium analysis, or something else, one may still ask where its coordinative labour is placed and by what mechanism it is performed. Indeed Hodgson’s observation that markets remain institutions even through disruptive disequilibria underscores this taxonomy’s neutrality: that particular institution can persist through an arbitrary sequence of equilibria and disequilibria. The taxonomy therefore cannot be utilized to settle the question of what institutions are fundamentally, but may be utilized to argue that a variety of patterns involving rules and placement can characterize different features of institutional order.

The profession is an especially instructive layered case. As Abbott’s (Reference Abbott1988) ecology of jurisdictions, surveyed in the section Competing accounts and their limits, already suggests, its differentiated unit is the role: physician, attorney, architect, and the like are person-shaped positions that supply standing bases for allocating responsibilities and jurisdiction. Yet ongoing integration can proceed diffusely through professional norms and reputation, whereas entry into the field itself and jurisdiction issues may be intermediated through licensing boards or professional bodies. The mechanisms governing who may enter need not be the mechanisms coordinating those already in the role.

Table 1 summarizes the three placement families and the two principal mechanisms distinguished within each.

Table 1.

Placement families and mechanisms of integration

Table 1. Long description

A table categorizing coordination mechanisms into placement families. The table has four rows and four columns. Column headers are Placement, Mechanism, How coordination is accomplished, and Paradigm case. Row labels are Non-mediated, Intermediated, and Diffuse. Row 1: Non-mediated, Mutual adjustment, Participants orient and respond directly during performance, Bucket brigade; string quartet. Row 2: Non-mediated, Negotiated coordination, Participants settle an allocation, plan, sequence, or terms among themselves, Hikers dividing loads; collaborators dividing a project. Row 3: Intermediated, Direction, A coordinator exercises discretionary judgement, Emergent leader; foreman. Row 4: Intermediated, Administered rule, An organization coordinates through standing offices, procedures, and rules, Bureaucracy; firm; hospital. Row 5: Diffuse, Signal-mediated coupling, Distributed signals coordinate responses without a coordinator, Market price; spot labour market. Row 6: Diffuse, Sedimented normative coordination, Socially available norms, conventions, classifications, expectations, and precedents shape and align conduct, Caste; gendered care work; professional custom.

Coordinative labour, once assigned, may therefore itself be divided: one coordinator may distribute parts of that work among several offices, or participants may retain some coordinative work while assigning other parts elsewhere.

Vigilance and revising labour

The principle of placement applies not only to the labour of coordinating an arrangement but also to the labour of watchfully keeping its terms under review. A division of labour may call for labour to maintain the terms on which coordination proceeds, and those terms may on occasion be contested or require revision. That work need not be undertaken by the same entities or units that implement the original scheme. A bureaucracy may coordinate through administered rules and also reserve revision to designated offices. But the prerogative to revise or contest these same rules may, in a civic setting, reside with the governed as a whole.

Consider a university department we will just refer to as Department D. Suppose it governs much of its ordinary life under practices widely approved by the profession – practices of shared governance and academic freedom – in connection with curriculum, hiring, and promotion policies – while remaining nested within a central administration that can direct it from above. If that administration presses Department D to alter its practices in order to shift the university’s profile, the department may take refuge in bylaws enshrining shared governance that record the terms under which its members understand their common work to be governed.

Those bylaws do not eliminate intermediated coordination: chairs, committees, standing procedures, and university rules still organize much of ordinary departmental life. But when pressure from above mounts, members can invoke the bylaws against encroachment, demand justification for departures from them, and, if necessary, use those bylaws as a basis for contesting or seeking renegotiation of administration-imposed strictures. The ongoing coordination is largely intermediated through offices, committees, and administered rules, but the labour of keeping its terms contestable and revisable need not be.

Call the standing distributed capacity that keeps such revision available vigilance. Vigilance includes noticing encroachment or malfunction, contesting departures from the terms of coordination, demanding justification, and activating revision when required; it need not mean continual rule-writing. What is distinctively civic is that vigilance and revising labour may be performed by the governed body even when everyday coordination is administered through offices and rules. Ostrom’s (Reference Ostrom1990) common-pool institutions exemplify arrangements in which a defined body of participants and collective-choice procedures help retain this revising capacity among those governed.

Relations of the axes

The two axes answer different questions and therefore cross-cut. The primary axis asks what kind of unit bears the differentiated assignment; the secondary asks how the coordinative labour is placed and by what mechanism it operates. A house-moving task can be coordinated non-mediatedly among friends, intermediated through a foreman, or allocated diffusely through spot labour contracting. A role-differentiated arrangement can be coordinated by mutual adjustment, as in a quartet with person-shaped roles, by administered rule, as in a hospital, or diffusely through professional norms. A station-differentiated order can be enacted directly in local relations, as when gendered expectations in a boss/secretary or doctor/nurse relation are sustained in the participants’ own interaction; it can be administered through law and offices, or reproduced diffusely through inherited expectation. The source and scale of an assignment need not coincide with the scale at which its performance is integrated.

The spot labour market is especially useful because it fills a combination that can otherwise look awkward: task differentiation with diffuse integration. The same first-order unit can therefore be coordinated in all three placement families. Equally, station differentiation need not be diffuse: a legally administered status order is station-differentiated on the primary axis and intermediated on the secondary. These off-diagonal cases establish the independence of the axes more strongly than the paradigm cases alone. The heterogeneous literatures surveyed in the section Competing accounts and their limits can thus be seen to isolate different variables or regions of this space rather than to offer mutually exclusive descriptions of a single structural dimension.

The taxonomy among its rivals

Three existing frameworks come nearest to different regions of the account. Functional differentiation (Luhmann, Reference Luhmann1982; Parsons, Reference Parsons1971) distinguishes social subsystems by the function they discharge, whereas the present primary axis distinguishes the unit of assignment and the secondary axis the distribution of coordinative labour. Polanyi’s (Reference Polanyi1944, 1957) reciprocity, redistribution, and exchange map naturally onto non-mediated, intermediated, and diffuse integration respectively, whereas the present scheme distinguishes mechanisms within those placements and reaches beyond economic exchange to offices, professions, civic bodies, and inherited normative orders. Treatments of divided cognitive labour likewise illuminate particular mechanisms rather than the full space (Goldberg, Reference Goldberg2011; Kitcher, Reference Kitcher1990). The gain is not that placement supersedes these frameworks but that it locates their explananda within a common taxonomy.

Selection and movement across the space

A clean division of theoretical labour between taxonomic distinctions and causal-historical explanation (that may invoke mechanisms) allows us to specify more precisely what an account of organizational change has to explain. Because the axes cross-cut, a historical change may involve movement on the secondary axis, the primary axis, or both.

The economic canon already identifies forces capable of producing such movements. Smithian gains from specialization create reasons to divide labour, whereas coordination costs constrain the arrangements through which those gains can be secured (Becker and Murphy, Reference Becker and Murphy1992; Smith, Reference Smith1776; Stigler, Reference Stigler1951).

The gains from specialization are not equally realizable across all structural arrangements. Smith’s familiar gains from repetition and reduced switching time, and Babbage’s further principle that division permits each process to draw only on the degree of skill it requires, depend on conditions the taxonomy leaves open rather than guarantees (Babbage, Reference Babbage1832). Repetition requires continuity of assignment; saving transfer time favours proximity or cheap handoff; specialized machinery involves indivisibilities and capital commitments; the Babbage principle presupposes labour markets capable of separately pricing differentiated skill. These conditions can make some pairings of differentiation and placement more economical than others without turning any of them into defining features of the forms themselves.

Coase (Reference Coase1937) and Williamson (Reference Williamson1975, Reference Williamson1985) explain why some transactions are organized within firms rather than through markets, whereas Langlois (Reference Langlois2003) follows movement across that boundary with capabilities and shifts in historical conditions. Scale, knowledge, uncertainty, observability, communication technology, and transaction frequency can alter the comparative attractiveness of available arrangements. Our taxonomy identifies structural differences among those arrangements without foreclosing the choice of any on a given occasion, while at the same time providing a view of the landscape across which choices can be traced.

A discrete task may be purchased on a spot labour market and integrated into a larger enterprise through prices and bids, or the same task may be brought within a firm and coordinated through direction or administered rule. That is a change on the secondary axis without any necessary change on the primary. Separately, an organization may constitute a person-shaped role that governs where future tasks, permissions, and responsibilities land. That is a change in the unit of differentiated assignment. Economic explanations of organizational change should therefore not assume that movement in integration and movement in differentiation co-occur, even when both are observed in the same historical episode.

Recent economic analyses suggest new tracings. Knowledge and communication costs make the gains/costs comparison more discriminating. According to Garicano (Reference Garicano2000), costly knowledge acquisition and communication produce knowledge hierarchies: routine problems are handled with locally available knowledge, whereas harder problems are referred to specialists. In the present terms, the model bears first on the constitution and use of differentiated roles – and on the routing of tasks among them – within an intermediated organization. It therefore need not represent movement between placement families at all. Bloom et al. (Reference Bloom, Garicano, Sadun and Van Reenen2014) show, moreover, that information and communication technologies can pull organizational authority in opposite directions: better information technologies are associated with greater autonomy for workers and plant managers and a wider managerial span of control, whereas technologies that lower communication costs are associated with less autonomy. These results likewise need not imply movement between placement families. They can alter the distribution of decision rights and, where those decisions are coordinative, the seat of coordinative labour within an intermediated organization – pushing discretion downward towards workers or plant managers, or making direction from higher-level coordinators easier to sustain. Autonomy within a hierarchy is therefore not by itself non-mediated integration. The taxonomy lets us distinguish redistribution within a placement family from movement between placement families.

The tasks/jobs literature must be brought into contact with the primary axis with care, because its categories do not align exactly with our own. In that literature, jobs and occupations are commonly represented as bundles of tasks; for us, by contrast, a role is not a bundle of tasks but a person-shaped standing position that allocates future tasks, permissions, responsibilities, and authority. The mismatch is analytically useful rather than merely a caution: task-based models can register churn in a job’s task content without, by themselves, distinguishing that churn from persistence or reconstitution of the standing role that bears it. That literature suggests that automation can displace labour from some tasks, whereas new tasks create new domains for it (Acemoglu and Autor, Reference Acemoglu, Autor, Ashenfelter and Card2011; Acemoglu and Restrepo, Reference Acemoglu and Restrepo2019), and new occupational specialities continue to emerge (Autor et al., Reference Autor, Chin, Salomons and Seegmiller2024). Task content can therefore change without constituting or abolishing a role in our sense, whereas a person-shaped role can survive substantial change in its task bundle.

Entrepreneurial judgement supplies another source of movement through the space. Foss and Klein (Reference Foss and Klein2015), drawing on Knight (Reference Knight1921), model entrepreneurs as resource owners who exercise judgement under uncertainty by controlling, combining, and deploying heterogeneous assets in anticipation of profit. Entrepreneurship is not a category in our taxonomy, but one source of movements across the terrain carved out by it. Among the judgements entrepreneurs make, in our terms, are judgements about placement itself: whether coordination is left to market signals, retained among first-order parties, or brought inside an organization under direction or administered rule. Entrepreneurial judgement may bear on the primary axis as well: an entrepreneur may redesign the decomposition of labour without changing its unit of assignment, or may change that unit by constituting a standing role for contributions previously allocated as discrete tasks.

Generative AI renders separation of landscape from movement across it especially visible. Brynjolfsson et al. (Reference Brynjolfsson, Li and Raymond2025) find that an AI assistant raised customer-support productivity by 15 percent on average, with larger gains for less experienced and lower-skilled workers and evidence of accelerated learning. A technology that makes experienced guidance cheaply available changes knowledge-transmission and escalation costs. It can reallocate tasks while roles persist, alter role content or jurisdiction, or redistribute coordinative judgement within an intermediated organization by reducing the need to escalate routine problems to supervisors. If adoption also changes whether work is obtained through market contracting or brought inside the firm, it can alter placement as well. AI adoption therefore need not represent a single kind of organizational change, but may affect task reallocation, role change, within-placement redistribution, and movement of labour between placement families.

Requirements for persistence

Each placement of coordinative labour also imposes characteristic requirements on persistence. These identify conditions that must continue to be met if coordination is to remain operative, without implying that failure determines any particular successor arrangement. Non-mediated coordination requires continuing coordinative activity by the participants themselves. Mutual adjustment depends on sustained reciprocal attention; negotiated coordination depends on the continuing efficacy of the settlement reached. These mechanisms do not require a separate organizational apparatus, but they do require attention, good communication skills, and willingness to coordinate. When these fail, the relevant vulnerability is dissipation or breakdown of the settlement, not a necessary transition to some other placement. The end of an episode does not, however, prevent re-instantiation: the same task-divided scheme can be assembled again under the guidance of precedent without a standing role structure persisting between episodes.

Intermediated coordination depends instead on the performance of an identifiable coordinator or organization. Direction requires discretionary judgement to remain effective; administered rule requires offices, procedures, and channels of implementation to continue doing their work. Such arrangements can support coordination across scale and time, but their characteristic vulnerabilities include entrenchment of discretion, ossification or proliferation of rules, and redirection of organizational capacities towards ends other than those for which the coordination is maintained. None of these possibilities implies that direction naturally becomes office or that offices naturally expand; they identify failure modes that become available once coordinative labour is intermediated.

Diffuse mechanisms pose different requirements. Signal-mediated coupling depends on the continued production, transmission, and uptake of usable signals; when signals become noisy, distorted, or disconnected from relevant activity, coordination deteriorates. Because no coordinator controls the mechanism as a whole, deliberate redesign ordinarily works by altering the institutional arrangements through which signals are produced and transmitted – as when a regulator, exchange, platform, or central bank changes the environment in which market signals operate.

Sedimented normative coordination depends on the continued social availability and reproduction of norms, classifications, expectations, and precedents. Processes of social closure and social reproduction can help explain how particular station arrangements are maintained. Closure can restrict access to favoured stations through licensing, credentialing, certification, union control, law, or other devices (Weeden and Grusky, Reference Weeden and Grusky2005; Weeden, Reference Weeden2002); reproduction can naturalize assignments across generations, as in gendered unpaid domestic and caring labour (Federici, Reference Federici2004; Folbre, Reference Folbre2001; Kittay, Reference Kittay1999). These processes help explain the durability of particular station arrangements without defining station differentiation or diffuse integration itself. Their significance is causal and historical: they identify processes by which a structurally described arrangement is maintained. And should they face erosion, that has consequences for all the arrangements that rely on their persistence.

Civic arrangements are examples where vigilance and revising labour must be tracked separately from ordinary coordination. Everyday work may remain heavily rule-governed and intermediated whereas the governed retain practical power to scrutinize and reopen those rules. Persistence of self-government therefore depends not simply on the continued existence of participatory procedures but on whether effective vigilance and revising labour remain distributed among the governed. Participatory procedures can therefore remain intact even when effective revising capacity has shifted elsewhere.

Institutional conditions

The comparative gains and costs that shape selection and the requirements that sustain an arrangement once selected are institution-dependent. Contract enforcement and property arrangements affect market coordination; licensing and professional jurisdiction constrain role occupancy; collective-choice procedures affect where vigilance and revising labour remain; inherited norms affect which assignments are socially available or costly to contest. Such conditions are further parameters on selection and persistence. The same change in specialization gains or technology can therefore support different arrangements under different institutional conditions (Ostrom, Reference Ostrom1990; Weeden, Reference Weeden2002; Williamson, Reference Williamson1985).

These are strictly requirements and vulnerabilities. Placement does not dictate a form’s origin, duration, or successor. It identifies what coordinative work must be accomplished, where that work is socially distributed, and some characteristic ways it can become unavailable or ineffective. The gains from a division of labour, the costs of coordinating it, and the institutional and technological conditions under which those gains and costs are realized explain which arrangements are selected and how they change.