Last updated: 2026-10-10

U
Undergraduate level

How Does an Organisation Retrieve What It Knows?

A new hire with a question spends twenty minutes searching a wiki, finds three contradictory pages, and gives up. An experienced colleague with the same question sends one message to one person and has an answer in five minutes. Nothing about the organisation's stored knowledge changed between those two attempts — what changed was which retrieval mechanism each person had access to. Here we look at those mechanisms, since the previous page's question of where knowledge lives only matters practically once something has to be found again.

A formal search — a keyword query against a wiki, a database lookup — is one retrieval mechanism among several, and often not the fastest one available. This cluster's page on organisational roles already covers the deeper mechanism: Wegner's transactive memory, a distributed, shared answer to who in the group is likely to know what, routed by trust and role rather than by keyword match. Where a formal search returns whatever matches the query regardless of who's asking, transactive-memory-based retrieval is personal — the new hire and the experienced colleague are drawing on genuinely different retrieval systems, not the same system with different skill levels applied to it, because the colleague's system includes a model of which specific person to ask that the wiki search doesn't have access to at all.

Cues, Prompts, and Organisational Attention FoundationalKnowledge that endures for decades — core principles

Peter Pirolli and Stuart Card's information foraging theory treats information-seeking as governed by information scent — cues in the environment that predict whether following a particular path is likely to lead somewhere useful, the same way a forager reads indirect signs of where food is likely to be rather than searching exhaustively1. An organisation runs on the same logic at a larger scale. An anomaly in a report, a familiar phrase in an email, a recurring item on a meeting agenda, a dashboard crossing a threshold — these are an organisation's information scent, the cues that trigger a retrieval episode rather than letting a relevant fact sit unused. This site's page on retrieval-augmented generation covers the equivalent, deliberately engineered mechanism for a language model: a query triggers a retrieval step rather than requiring everything relevant to already be loaded in context. The parallel is structural, not literal — an organisation's retrieval cue is a person noticing something, not a vector-similarity search — but both solve the identical shaped problem: don't try to keep everything active at once, retrieve what's needed when something signals it's needed.

Attention as a Scarce Resource Applied / MethodologicalKnowledge with a 5–10 year half-life — stable practice

Information scent only helps if someone is positioned to notice it, and who's positioned to notice what is exactly the question this cluster's page on roles answers with its exploratory and ethnographic functions. An organisation with no one currently occupying an exploratory role toward a given class of anomaly won't retrieve the relevant knowledge even if it's sitting in a document three people could point to — the cue arrives, and nobody with the right attention is there to read it as a cue at all. Retrieval failure, looked at this way, is often an attention failure wearing the costume of a missing document.

Strongest Objection: Retrieval Paths Encode Power, Not Just Efficiency FoundationalKnowledge that endures for decades — core principles

Being routed to the fastest answer is not neutral. Whose question gets a same-day reply from a senior colleague and whose gets forwarded twice before anyone responds often tracks seniority, network position, and who's already trusted — not simply who's actually best placed to answer. A retrieval system built entirely on personal trust networks will reliably reproduce whatever pattern of access and exclusion already exists in those networks, faster and less visibly than a formal, auditable search ever could. This page treats that as a flag rather than a full argument — a later page in this part takes up directly who actually gets heard when an organisation claims to know something.

Provisional Conclusion FoundationalKnowledge that endures for decades — core principles

An organisation retrieves what it knows through a layered combination of formal search, trust-routed transactive memory, and attention cued by whoever currently occupies a noticing role — not through any single mechanism, and rarely through the one written into the induction manual. The fastest retrieval path and the fairest one are not guaranteed to be the same path — retrieval speed alone isn't evidence that an organisation's knowledge system is working well.

Questions for Further Thought

  • Could an organisation measure its own information scent deliberately — tracking which cues reliably trigger useful retrieval and which get missed — the way a search system's designers measure query success?
  • Is there a retrieval mechanism that could preserve the speed of a trust-routed question without inheriting the access patterns of whatever personal network the question happens to travel through?
  • Does retrieval speed itself change what gets asked — are some genuinely important but slow-to-answer questions quietly avoided because the organisation has learned that no fast path to an answer exists?

Further Reading

  • Pirolli, P., & Card, S. (1999). Information foraging. Psychological Review, 106(4), 643–675.

References

  1. Pirolli, P., & Card, S. (1999). Information foraging. Psychological Review, 106(4), 643–675. ↩