A document store is not a knowledge management system
Shelving documents in a repository is storage, not knowledge management; the presence of the repository often produces false confidence that the problem is solved.
Most mid-sized organisations, asked whether their knowledge is managed, say yes. They point to SharePoint, or Drive, or Notion, or a document management system the firm paid for. From a distance, it looks like the knowledge is in order.
Look closer and the picture often changes. Repositories fill up with documents no one maintains. Version control is nominal. There are multiple copies of everything and no one can reliably say which is current. Critical context lives in email threads, in the heads of senior staff, in informal practices nobody has written down. What the organisation calls a knowledge management system is, in practice, a document storage system — and not always a good one.
Why the confusion matters
False confidence is worse than acknowledged absence. A firm that knows its knowledge is not managed will, at least sometimes, invest in fixing the problem. A firm that believes it has already solved the problem will not — and will instead layer AI tools on top of a foundation that cannot support them, producing the plateau described in A tools-first AI rollout that plateaued.
The diagnostic value of the distinction comes from asking the right question. Not “where are our documents?” but “where does the knowledge live, and how would we know?”. The second question tends to surface uncomfortable answers: in one partner’s head, in a draft that never got filed, in a client-specific email from eighteen months ago, in a practice that nobody has codified but everybody more or less follows.
The aim: authoritative, not comprehensive
A specific consequence follows from treating a document store as a knowledge management system. The implicit goal becomes completeness — everything the firm has ever produced, retained somewhere findable. In the AI era, that goal is actively counterproductive. A knowledge base optimised for AI consumption aims at the opposite: authoritative, current, mutually consistent material, with active pruning of anything that dilutes the rest. See Useful AI is a context problem for why dilution from over-inclusion damages AI output.
The practical discipline is pruning. Archiving outdated materials, deleting redundant documents, resolving conflicting policies into single positions. Every irrelevant document, every outdated procedure, every conflicting guideline dilutes the influence of accurate, current knowledge. Pruning is unglamorous and ongoing; it does not earn a project budget and does not show up on anyone’s KPIs. It is also what separates a knowledge management system from a document store in practice.
How to apply the check
A short set of tests is usually enough to establish whether an organisation is looking at storage or at knowledge management.
Can a major client’s context be reconstructed, in a day, if the lead partner left tomorrow? Is the current version of a given document unambiguous from where it is stored, or does it require institutional memory to identify? When senior staff finish a complex engagement, is there a process that captures what they learned, or does the knowledge leave with them? Are the documents structured in a form that another person — or an AI tool — could usefully consume without a tour guide?
Three further questions test the governance layer that supports any of this. Who is accountable for document or knowledge management in your organisation? What is your process for archiving out-of-date documents? Do you have a taxonomy or standardised schema for document metadata? “I don’t know” or “no single person” answers to any of these are themselves a diagnosis: a knowledge management system without clear ownership, active archival discipline and a metadata schema is not operating as a knowledge management system even if it is widely understood to be one.
If any of those questions is uncomfortable, the organisation has a storage system, not a knowledge management system. That is the honest starting point. The work described in Start with knowledge management, not tools follows.