One of the more puzzling experiences inside a large institution occurs when a team spends considerable time trying to solve a problem, perhaps commissioning new analysis, searching through documents or looking outside the organisation for expertise, only to discover eventually that somebody elsewhere in the same institution had already encountered almost exactly the same question. The answer, or at least an important part of it, was available all along, but the people who needed it did not know where to look and the people who possessed it had no reason to realise that their knowledge had become relevant elsewhere. What initially appeared to be a knowledge gap was therefore something rather different: the institution possessed knowledge that it was unable to make available to itself.

This possibility complicates the apparently simple question of what an institution knows, because knowledge can exist at several levels at once. An individual may possess expertise, a team may have accumulated experience, a regional office may have solved a recurring problem or a document may contain exactly the analysis required for a new decision, while the institution as a whole continues to behave as though none of that knowledge existed. The important distinction is therefore not merely between knowing and not knowing, but between possession and availability. Knowledge can be present somewhere within an organisation without being institutionally available at the moment and place where it could influence action.

Large institutions are particularly vulnerable to this condition because knowledge naturally becomes distributed as organisations grow more complex. Expertise develops within professions, departments, programmes, geographical units and specialist communities, each of which becomes increasingly capable of understanding its own part of institutional reality. This division of cognitive labour is indispensable: no individual could plausibly master everything that a modern public institution needs to know. Yet the same distribution that makes sophisticated institutional knowledge possible also creates a problem of discoverability. The more places in which knowledge can reside, the less plausible it becomes that any particular person will know where all relevant knowledge is located.

The familiar reassurance that “somebody in the organisation will know” can consequently be misleading. Somebody knowing something does not automatically make that knowledge an institutional capability, because the value of expertise depends partly on whether the institution can connect it with the problems to which it applies. An engineer may understand a technical constraint that a policy team has never encountered; a frontline employee may recognise a failure pattern invisible in central reporting; a regional office may have developed a practical solution that another region is now trying to invent independently. In each case the knowledge exists, but its institutional usefulness depends on a connection that may or may not occur.

This is why knowledge location should be understood as part of institutional cognition rather than as a merely administrative convenience. An institution does not need a perfect inventory of everything every person knows, which would in any case be impossible to maintain, but it does need some capacity to form an imperfect map of its own cognitive resources. When an unfamiliar problem appears, the organisation should have ways of asking not only what information is already stored, but who may understand the problem, where something similar may have happened before, which unit has relevant experience and through which relationships a question might travel far enough to encounter expertise that is not already visible to those asking it.

The problem is therefore partly one of routing. Knowledge may be excellent on one side of an institutional boundary and the need for that knowledge may be perfectly clear on the other, yet nothing cognitively useful happens unless the two can meet. A library containing precisely the book a reader needs, but lacking any catalogue, classification system or other way of locating it, illustrates the basic difficulty: physical possession of the book does not guarantee practical access to its contents. Human expertise can become institutionally inaccessible in much the same way, particularly when people know more than their formal job descriptions reveal or when the relevance of experience becomes apparent only after a new problem has emerged.

This also explains why simply adding more expertise does not necessarily solve an institution’s knowledge problems. An organisation can employ increasingly capable specialists while becoming progressively less aware of what those specialists collectively know. Indeed, deeper expertise can sometimes make discoverability harder because specialists develop vocabularies, categories and problem definitions that outsiders may not recognise. Someone searching for help may therefore fail to identify the relevant expert not because the expertise is hidden deliberately, but because the problem and the knowledge capable of addressing it are described in different cognitive languages.

Institutions use many mechanisms to reduce this gap, from professional networks and communities of practice to searchable repositories, expertise directories, cross-unit forums and knowledge-brokering roles. Each can help, but none should be confused with the capability itself. A directory may accurately describe formal responsibilities while revealing little about accumulated experience; a repository may contain thousands of relevant documents while giving users no reliable way to distinguish the useful ones; informal networks may route questions remarkably effectively while depending on a handful of unusually well-connected individuals. What matters is not the existence of a particular tool, but whether the institution has enough architecture for knowledge and problems to encounter one another with reasonable reliability.

Digital technologies and artificial intelligence can expand this capability by making recorded knowledge easier to search and by identifying relationships between a new question and earlier documents, cases or areas of expertise that would otherwise remain obscure. Yet these technologies also make the distinction between possession and availability more important rather than less. A search system can work only with the traces it can access, while valuable tacit experience may remain invisible, poorly documented work may never enter the searchable environment and formal organisational metadata may provide an incomplete picture of where practical expertise actually resides. A powerful search interface can even produce false confidence if users begin to assume that knowledge which cannot be retrieved therefore does not exist.

Discoverability also works in more than one direction. We normally imagine a person with a problem searching for someone who knows the answer, but institutions can also create conditions in which knowledge becomes more capable of finding the problems to which it is relevant. Cross-unit conversations, shared problem registers and mechanisms for exposing emerging challenges allow people to recognise that an experience or expertise developed in one context has unexpectedly become useful in another. Institutional availability is strengthened when both movements are possible: when problems can find knowledge and when knowledge can recognise problems.

Seen in this way, the question “does the institution know?” becomes less straightforward than it initially appears. Before concluding that new knowledge must be created, commissioned or acquired, it may be necessary to ask whether the knowledge genuinely does not exist or whether the institution has simply failed to discover that it already possesses it. Confusing these situations can lead organisations to repeat analysis they have already performed, recruit expertise they already employ or solve problems that another part of the institution has effectively solved before. The resulting waste is not merely financial or administrative; it reveals an institution whose usable intelligence is smaller than the sum of the knowledge contained within it.

A knowledgeable institution, then, is not simply an organisation populated by knowledgeable people. It is an organisation in which relevant knowledge has a reasonable chance of becoming available beyond the place where it happens to reside, because the architecture provides enough awareness, search, connection and routing for distributed expertise to become institutionally consequential. The distance between “someone here knows” and “the institution knows” is partly the distance between possession and availability: until knowledge can encounter the problem that needs it, an institution may contain the answer while continuing to experience the world as though the answer were nowhere to be found.