Hierarchies solve an essential institutional problem: they establish who is authorised to decide. In a complex organisation, not every question can be negotiated among everyone affected by it, nor can responsibility remain permanently distributed without a point at which judgement becomes consequential. Institutions therefore create vertical structures through which authority is delegated, decisions are escalated, responsibilities are assigned and accountability can ultimately be located. These structures make coordinated action possible at scales that would otherwise be extraordinarily difficult to sustain. Yet the architecture that works well for moving authority through an institution is not necessarily the architecture through which knowledge needs to move. AUTHORITY FLOW ≠ KNOWLEDGE FLOW, and many institutional failures emerge when organisations assume that information and understanding should travel through the same vertical channels that organise formal power.

The distinction matters because authority and knowledge have different organisational properties. Authority can be deliberately allocated: a minister possesses powers that an analyst does not, a director can approve decisions that a frontline employee cannot, and a regulatory body can impose requirements that an advisory team can only recommend. Knowledge is less obedient to organisational design. Relevant understanding may emerge anywhere in the institution, including places with very little formal authority. Frontline staff may recognise a change before senior management does; analysts may identify patterns that operational teams cannot see; local offices may understand contextual variation invisible from headquarters; and people outside the institution may possess knowledge unavailable through any internal reporting system. The location of authority can therefore be designed relatively explicitly, while the location of knowledge is partly determined by where experience, expertise and observation happen to occur.

A hierarchy becomes cognitively problematic when the institution assumes that knowledge should travel upward through the same chain through which authority travels downward. In principle, this arrangement appears orderly: information moves from operational levels towards decision-makers, decisions move back towards implementation, and each managerial layer performs a filtering and coordinating function. In practice, however, knowledge can change as it moves. Detail may be compressed because senior levels cannot process everything generated below them; uncertainty may disappear as ambiguous observations are converted into reportable categories; disagreement may be resolved before decision-makers can see that it existed; and weak signals may be discarded because they do not yet satisfy established thresholds for escalation. By the time information reaches the point of formal authority, it can be more manageable and less informative at the same time.

Some filtering is unavoidable. Senior decision-makers cannot receive every observation made throughout a large institution, and an organisation that attempted to transmit everything without selection would replace information scarcity with cognitive overload. The problem is therefore not that hierarchies filter knowledge, but that the criteria through which they filter it are often shaped by administrative requirements rather than cognitive ones. Information that matters for accountability, compliance or performance reporting may travel efficiently because formal systems are designed to carry it, while contextual knowledge, anomalies and emerging concerns may struggle to move because they do not fit established reporting categories. The hierarchy can consequently become excellent at transmitting what it already knows it needs to know while remaining comparatively weak at carrying information whose importance has not yet been recognised.

This asymmetry is particularly visible during emerging problems. Early warnings rarely arrive as complete diagnoses. They appear as unusual cases, local inconsistencies, small deviations, complaints, professional intuitions or patterns whose significance is uncertain. People close to the problem may sense that something is changing without yet possessing the evidence required to formulate a convincing institutional claim. Hierarchical escalation, however, often demands increasing levels of certainty as information moves upwards. Each level reasonably asks whether the issue is sufficiently important to warrant senior attention, but the cumulative effect can be that uncertain information must become certain before it reaches the people capable of acting on uncertainty. An institution can therefore possess early knowledge of a developing problem while its formal decision architecture remains unaware of it.

The opposite problem can occur when authority moves downward. Senior leadership may identify a strategic objective and communicate it clearly through the hierarchy, yet the knowledge required to translate that objective into intelligent action may be distributed across operational contexts that differ substantially from one another. A uniform instruction can travel efficiently because the authority behind it is clear, while the contextual knowledge required to adapt it intelligently may have no equally effective route back into decision-making. The institution then becomes asymmetrical: decisions can move downward faster than learning can move upward. What appears from the centre as inconsistent implementation may sometimes be the result of local actors encountering realities that the original decision architecture could not adequately represent.

None of this means that institutions should abandon hierarchy in favour of unrestricted networks. Hierarchies remain essential for accountability, legal authority, resource allocation and the resolution of disputes that cannot remain indefinitely open. Networks can move knowledge laterally and connect expertise across boundaries, but they can also make responsibility ambiguous and leave collective judgement without a legitimate point of decision. The challenge is therefore not to choose between hierarchy and network as competing organisational ideals. It is to recognise that the architecture of authority and the architecture of knowledge solve different institutional problems, and an intelligent organisation may need them to overlap in some places while diverging deliberately in others.

This becomes especially important in highly specialised institutions. Expertise rarely follows the organisational chart neatly because the knowledge required for a decision may cut across several professional and departmental boundaries. A senior official may possess formal responsibility for an issue while relying on economists, lawyers, scientists, operational staff and local practitioners whose knowledge cannot be reduced to a single vertical reporting chain. If every contribution must travel upward through its own hierarchy before meeting the others near the top, integration occurs late and after considerable filtering. More direct lateral relationships can sometimes allow different forms of expertise to encounter one another earlier, improving the quality of the judgement that eventually reaches formal authority without changing who is ultimately authorised to decide.

Digital systems can alter this geometry because information no longer needs to travel physically through managerial layers in order to become accessible elsewhere in an organisation. Shared platforms, dashboards and interoperable data can allow senior leaders, analysts and operational teams to observe information simultaneously. Yet digital accessibility does not eliminate the distinction between authority and knowledge. The existence of a dashboard does not determine which signals deserve attention, whose interpretation should matter or whether an anomaly can challenge an established judgement. Indeed, digital systems can reproduce hierarchical assumptions in new forms if access permissions, metrics and reporting structures are designed primarily around existing chains of command. Technology can shorten the route travelled by information while leaving unchanged the institutional rules governing whether that information becomes consequential.

Artificial intelligence extends the same challenge. AI systems can identify patterns across organisational levels, aggregate dispersed information and surface anomalies that might previously have disappeared inside reporting chains. This creates important opportunities for institutional cognition, but it also raises questions about where model-generated knowledge enters authority structures. A system may identify a significant risk without there being a clear institutional pathway through which that finding can challenge a decision already supported by senior leadership. Conversely, decision-makers may rely heavily on AI-generated analysis while becoming more distant from contextual knowledge held by people close to implementation. The cognitive value of AI therefore depends partly on whether institutions can connect new forms of knowledge with authority without assuming that the two should occupy the same organisational location.

The distinction also changes how leaders might understand their own role. Seniority often brings broader responsibility but not necessarily greater proximity to every relevant form of knowledge. A mature hierarchy would recognise this explicitly rather than treating the upward movement of authority as evidence that understanding has also become concentrated at the top. Leadership can then be conceived partly as the design of conditions under which knowledge can reach consequential judgement without requiring every knowledgeable actor to acquire formal authority. The aim is not to weaken decision rights but to make those rights cognitively permeable enough to be informed by expertise, experience and evidence wherever they are located.

This is particularly important when bad news is involved. Hierarchical systems can unintentionally create incentives for information to become more reassuring as it moves upward because each managerial layer is partly responsible for the performance being reported. Problems may be reframed as manageable, uncertainty may be reduced and local concerns may be interpreted as exceptions rather than systemic signals. No deliberate concealment is required for this effect to occur; ordinary organisational incentives can gradually transform what decision-makers receive. A cognitively robust institution therefore needs ways for consequential knowledge to travel that do not depend exclusively on the willingness or ability of every intermediate layer to recognise its importance.

Understanding these differences allows institutions to ask a more useful question than whether they are too hierarchical. The relevant issue is whether their hierarchy is being asked to perform cognitive functions for which it was not designed. An organisation may have an entirely appropriate distribution of authority and still need additional routes through which expertise, uncertainty, dissent and emerging evidence can travel. Those routes do not have to undermine hierarchy. Properly designed, they can strengthen legitimate authority by improving the quality of the knowledge available when authority is exercised.

Institutional intelligence therefore does not require authority and knowledge to be organised identically. It requires the institution to understand the relationship between them. Authority must eventually become locatable because decisions have consequences and accountability matters, while knowledge must often remain distributed because experience and expertise are distributed. A hierarchy becomes cognitively stronger not when all knowledge is forced to travel through the chain of command, but when the institution can preserve clear authority while allowing relevant knowledge to reach judgement through whatever pathways the problem actually requires.