When an institution produces a poor outcome, one of the most natural responses is to look for the part that failed. Which department made the mistake? Which team missed the warning? Which procedure was not followed? Which manager should have acted differently? This way of investigating failure is understandable because institutions are divided into identifiable organisational units, each with responsibilities that can be inspected and evaluated. Yet some institutional failures resist this form of diagnosis because no single department is performing particularly badly. Each unit may be competent within its own mandate, each may possess relevant expertise, and each may follow its procedures correctly, while the institution as a whole still produces an inadequate result. In these cases, searching harder for the defective component can obscure the actual problem, because COMPONENT PERFORMANCE ≠ RELATIONAL PERFORMANCE: an institution can fail not because one of its parts is weak, but because the relationships through which those parts combine their knowledge and action are inadequate.

Consider a public problem that crosses several administrative domains. One department possesses detailed information about the affected population, another controls funding, a third administers the relevant service, a fourth holds regulatory authority and a fifth collects data that could reveal whether the intervention is working. None of these organisations necessarily lacks competence. Each may perform well against its own objectives and provide exactly what its formal mandate requires. Yet the public problem does not experience government through those organisational boundaries. If relevant information cannot move from the department that possesses it to the department whose decision depends upon it, if responsibilities intersect without producing joint judgement, or if different units interpret the same situation through incompatible categories, then competent components can combine into an unintelligent institutional response.

This possibility changes the unit of analysis. Conventional performance management often evaluates departments individually because administrative responsibility is organised that way: budgets belong to units, targets are assigned to units, leaders are accountable for units and audits examine whether units fulfilled their obligations. These practices are necessary, but they can create a systematic blind spot when outcomes depend on relationships rather than components. If every department can demonstrate satisfactory performance within its own boundary, an institution may conclude that the system is functioning even while citizens continue to encounter delays, contradictions or unresolved problems that arise precisely when responsibility passes from one organisational domain to another. What looks like a collection of successful departments from inside the organisation can therefore look like a dysfunctional institution from outside it.

The familiar language of “silos” captures part of this phenomenon, but it can also oversimplify it. Departments are not necessarily separated because their staff are territorial, unwilling to cooperate or insufficiently committed to whole-of-government outcomes. Institutional differentiation exists for good reasons. Specialisation allows expertise to develop, makes responsibility intelligible and enables complex problems to be divided into manageable domains. A tax authority should not become a health ministry, and a regulatory agency should not abandon specialised judgement simply in order to become more integrated with every other public body. The problem begins when specialisation creates dependencies that the institutional architecture does not adequately recognise. Two departments can each be doing exactly what they were designed to do while the relationship between them remains incapable of supporting the combined judgement that a cross-boundary problem requires.

This is why relational failure cannot be diagnosed simply by asking whether departments communicate. Information can travel frequently between units without producing meaningful coordination. Meetings can multiply, documents can be circulated and shared platforms can be created while different parts of the institution continue to interpret the problem differently or remain unable to influence one another’s decisions. Communication describes movement; relational intelligence requires something more demanding, because knowledge arriving from another part of the institution must be able to become relevant to judgement. A department may receive information that it cannot interpret, understand evidence that it has no authority to act upon, or participate in coordination processes whose incentives still reward local rather than collective optimisation. The relationship exists formally, yet it does not perform the cognitive work the institution requires from it.

Many policy failures can consequently emerge from individually rational behaviour. A department may optimise processing speed because that is how its performance is measured, while another tries to reduce error and therefore introduces additional checks. Each response makes sense within its own local objective, but together they may generate delays that neither department experiences as its responsibility. One agency may minimise financial risk by imposing stricter eligibility requirements while another is tasked with increasing access to the same service. Again, both may behave rationally within their mandates while the institutional relationship between those objectives remains unresolved. When such tensions are treated as failures of individual units, management interventions can make the problem worse by demanding even stronger local performance from components whose optimisation is already producing poor system-level results.

The same pattern appears in knowledge. Institutions frequently possess the information required to recognise a problem without possessing it in a form that can be combined effectively. One department sees operational anomalies, another understands the legal constraints, another has longitudinal data and another hears directly from affected communities. No individual unit has the whole picture, but neither does the institution automatically acquire that picture simply because all of its parts possess pieces of it. Institutional cognition is relational because understanding often has to be assembled across organisational boundaries. The quality of that understanding therefore depends not only on the knowledge held inside departments but on whether the relationships between them allow different forms of knowledge to encounter, challenge and modify one another.

This helps explain why adding expertise to an institution does not always improve its intelligence. If the relevant relational architecture remains unchanged, new expertise can simply create another specialised node whose knowledge struggles to affect the wider system. Governments can therefore accumulate analysts, specialist teams, data units and advisory functions while continuing to experience familiar coordination failures. The problem is not that the additional expertise has no value, but that the institution may be strengthening components without strengthening the relationships through which their capabilities become collective. A system composed of increasingly capable parts can remain limited by the architecture connecting those parts.

Digital integration does not automatically resolve this problem either. Shared databases, interoperable systems and common digital platforms can reduce important technical barriers, but relational performance is not identical to information accessibility. Two departments may have access to the same data while interpreting it differently, attaching different significance to the same indicators or possessing incompatible authority to respond. Technology can make information available across boundaries without determining how that information should alter judgement. Digital interoperability is therefore potentially valuable, but institutional coordination also depends on cognitive interoperability: whether different parts of the organisation can make one another’s knowledge meaningful enough to support combined action.

Relational failure can be particularly difficult to identify because responsibility becomes diffuse precisely where the problem is located. When a department fails internally, there is usually an identifiable locus for investigation and correction. When failure occurs between departments, each unit may be able to demonstrate that it fulfilled its own responsibilities. The missing function belongs to the relationship, yet relationships rarely have the same organisational visibility as departments. They may have no dedicated budget, leadership or performance indicators, and the work required to sustain them can appear secondary to the formal responsibilities of participating units. The institution can therefore possess elaborate systems for managing its components while having comparatively weak mechanisms for observing whether the connections between those components are performing adequately.

For citizens, this distinction is often much easier to perceive because their problems rarely respect administrative architecture. A person seeking support may move between agencies whose individual decisions are procedurally correct but collectively contradictory. A business may comply separately with several regulatory regimes whose combined requirements create effects that no regulator intended. A community may interact with multiple public bodies that each address one dimension of a problem while nobody is positioned to interpret how those dimensions interact. From the institutional perspective, these cases can look like complicated coordination challenges; from the outside, they reveal something more fundamental about cognition, because the institution is unable to assemble its distributed knowledge into an understanding proportionate to the problem.

Recognising relational failure does not imply that every cross-boundary problem requires more centralisation. Moving authority upward can sometimes improve coordination, but it can also separate decisions from specialised or local knowledge. Nor does the diagnosis imply that departments should dissolve their boundaries and work collectively on everything, which would sacrifice many of the benefits of specialisation. The relevant question is more precise: where does institutional performance depend on relationships that are currently weaker than the components they connect? Once that question is visible, coordination can be examined as an architectural property rather than treated merely as a matter of goodwill or communication.

This perspective also changes how institutional improvement can be evaluated. If the problem is relational, improving one department in isolation may have little effect even when that department becomes demonstrably more capable. The institution needs to observe what happens at interfaces: where information changes meaning, where responsibility passes between units, where different objectives interact, where knowledge and authority become separated and where no single actor can see enough of the system to understand the consequences of local decisions. These interfaces are not peripheral administrative details. In many complex institutions, they are among the places where collective cognition is either created or lost.

An institution can therefore contain highly capable departments and still behave unintelligently as a whole. The mistake is to assume that collective capability will emerge automatically from component capability, as though organisational competence were simply additive. In reality, institutional intelligence depends partly on what happens between specialised parts: whether their knowledge can meet, whether conflicting interpretations can become visible, whether dependencies can be recognised and whether local decisions can be understood in relation to wider consequences. Sometimes the most important question after institutional failure is not which department performed badly, but whether the institution ever created relationships capable of turning the competence of its separate departments into coherent collective judgement.