When an institution is described as fragmented, the phrase often evokes an image of organisational disorder. Responsibilities overlap, departments compete, information is scattered, authority is unclear and nobody seems certain who should act. Fragmentation appears to be another name for poor organisation. Yet institutions can experience a very different form of fragmentation while displaying almost none of these symptoms. Responsibilities can be precisely allocated, reporting lines can be clear, procedures can function reliably and specialist units can perform their individual tasks competently, while the institution as a whole remains unable to connect what its different parts know.

This possibility matters because organisational order and cognitive connectedness solve different problems. Organisational structure determines, among other things, who is responsible for what, where authority resides and how work is divided. Cognitive architecture concerns whether knowledge, interpretation and judgement produced in one part of that structure can become meaningfully available to another when institutional action requires it. A clear division of responsibility can support both, but it does not guarantee both. FORMAL ORGANISATIONAL ORDER ≠ COGNITIVE CONNECTEDNESS.

Indeed, specialisation can make the distinction more important. Complex institutions need specialised units because no single group can possess all the expertise required to govern complicated social, technical and economic systems. A health department should develop forms of knowledge that differ from those of a transport authority, a finance ministry or an environmental regulator. Within each organisation, further specialisation is equally necessary. Legal teams, operational units, analysts, inspectors and policy professionals attend to different dimensions of reality because institutional capability depends on cognitive differentiation.

The resulting institution may be exceptionally well organised. Each unit knows its mandate, maintains relevant expertise and performs defined functions. Yet the very success of this division of cognitive labour can produce boundaries across which knowledge becomes difficult to use. One department may understand a problem in terms that another does not recognise. The same evidence may acquire different institutional significance in different units. Shared words may conceal divergent concepts. Individually capable parts can therefore coexist with weak cognition at the level of the institution as a whole.

This is not necessarily a failure of communication. Messages may move efficiently between departments. Documents may be circulated, databases shared and meetings held regularly. Information can cross a boundary without the knowledge embedded in it becoming usable on the other side. A technical assessment may reach a policy team that lacks the conceptual framework needed to interpret its significance. An operational insight may be reported upward but lose the contextual knowledge that made it meaningful. An analysis produced for one decision environment may be received in another where its assumptions are no longer obvious.

The distinction between communication and cognitive connection is crucial. Communication asks whether something was transmitted. Cognitive interoperability asks whether what was transmitted can participate meaningfully in another part of the institution’s reasoning. A system can therefore possess abundant communication and remain cognitively fragmented.

Consider a government confronting an increase in extreme heat. Health authorities may understand the problem through mortality, hospital admissions and vulnerable populations. Urban planners may focus on building form, green space and neighbourhood design. Energy agencies may see pressure on electricity demand and grid resilience, while labour regulators examine workplace exposure and transport authorities consider infrastructure performance. This differentiation is useful because the problem genuinely has multiple dimensions. Yet if each unit’s knowledge remains enclosed within its own decision architecture, the government can become a collection of accurate partial understandings without developing an integrated institutional understanding of heat risk.

Nothing in this example requires administrative confusion. Each organisation can know exactly what it is supposed to do. The fragmentation lies elsewhere: the institution lacks sufficient mechanisms for the cognitive products of specialised units to modify one another. Health evidence does not reshape planning assumptions; infrastructure knowledge does not alter social-vulnerability analysis; energy scenarios do not enter decisions about urban form. The pieces coexist without becoming mutually consequential.

This reveals why clearer organisational charts cannot by themselves solve cognitive fragmentation. Reassigning responsibilities may sometimes be necessary, particularly when genuine gaps or overlaps exist. But if the underlying problem concerns incompatibility among forms of knowledge, changing boxes and reporting lines can leave the cognitive boundary intact. Two teams can be placed in the same department and still reason through mutually opaque frameworks. Conversely, units located in different organisations can become highly cognitively connected if they possess effective ways of translating and integrating their specialised knowledge.

Coordination is also insufficient if understood only as synchronising activity. Departments can coordinate deadlines, divide tasks and avoid duplication while continuing to interpret the underlying problem separately. Such coordination may be operationally excellent. It ensures that specialised actions do not interfere with one another. Cognitive integration asks a different question: can what one part learns alter what another part understands?

This is why COORDINATION ≠ INTEROPERABILITY. The distinction does not diminish the importance of coordination. Institutions need both. But coordinating separate actions is not equivalent to connecting the reasoning that generates those actions.

Fragmentation can become especially difficult to recognise when each specialised unit performs well according to its own criteria. Local competence can hide system-level weakness. If every department meets its objectives, there may be little reason to question the architecture connecting them. Yet public problems frequently cross the boundaries through which institutional responsibilities have been divided. Housing affects health; transport affects labour-market access; energy policy affects social vulnerability; education interacts with economic development. The more effectively each domain develops its own specialised cognition, the greater the need for mechanisms capable of reconnecting those forms of knowledge when the world refuses to respect the institution’s organisational boundaries.

The goal is not to erase those boundaries. A fully undifferentiated institution would lose the benefits of expertise and create an impossible integration burden. Nor does every piece of knowledge need to circulate everywhere. Cognitive connectedness is selective. What matters is whether relevant knowledge can cross a boundary when a decision depends on understanding relationships that the organisational structure has separated.

This requires more than centralisation. A central coordinating body can itself become another specialised node that receives information without fully absorbing the reasoning behind it. Similarly, creating a common database can improve informational access without resolving differences in meaning. The architecture has to support translation among forms of expertise, recognition of differing assumptions and integration where no single professional frame is sufficient.

Artificial intelligence could strengthen or weaken this capability. Shared analytical systems may create common informational resources across departments, helping previously disconnected units examine overlapping problems. But common tools do not guarantee common cognition. If different units interpret the same outputs through incompatible mandates and conceptual frameworks, a technically integrated platform can coexist with cognitive fragmentation. Alternatively, an AI system designed to translate among specialised representations might improve interoperability while preserving useful differences. The relevant question is not whether everyone uses the same technology, but whether knowledge produced in one part of the institution can become intelligible and consequential elsewhere.

This also distinguishes cognitive fragmentation from disagreement. A connected institution need not reach consensus. Two departments can understand each other’s reasoning perfectly and still recommend different actions because they represent legitimate competing objectives. In fact, genuine cognitive connection may make disagreements clearer by exposing the values, assumptions or trade-offs behind them. Fragmentation occurs when the institution cannot adequately bring those different forms of reasoning into relation at all.

Nor should connectedness be confused with uniformity. Institutional intelligence does not require every unit to think alike. A resilient cognitive architecture may depend precisely on maintaining multiple perspectives capable of challenging one another. The objective is not one institutional mind in which difference disappears, but an architecture in which differentiated knowledge can interact without first having to become identical.

Seen this way, fragmentation is not primarily about the number of departments an institution contains. A highly decentralised system can be cognitively connected, while a formally centralised organisation can be deeply fragmented. What matters is the quality of the interfaces between specialised forms of cognition. Can knowledge cross them? Can meanings be translated? Can different interpretations be compared? Can an insight generated in one domain change assumptions in another? Can the institution construct an integrated judgement when no single unit possesses enough of the problem to decide alone?

These questions reveal a dimension of institutional design that organisational charts rarely capture. Lines of authority tell us who reports to whom. Process maps tell us how work moves. Information systems tell us where records are stored and who can access them. None, by itself, tells us whether the institution’s differentiated knowledge can function as a connected cognitive system.

An institution can therefore be organised and fragmented at the same time. Its administrative architecture may successfully divide labour while its cognitive architecture fails to reconnect what that division has separated. The paradox is not evidence that specialisation was a mistake. It shows that specialisation creates an integration requirement alongside the expertise it makes possible.

The institution can be fragmented without being disorganised because organisational order and cognitive connectedness are different properties. Clear mandates, reliable procedures and effective specialised units can coexist with weak interfaces between the knowledge those units produce. The answer is not to eliminate differentiation or force every department into the same conceptual framework. It is to build forms of interoperability through which specialised knowledge can remain distinct while becoming mutually intelligible and consequential. Institutional intelligence depends not only on having capable parts, but on whether those parts can participate in cognition that no part could perform alone.