"Tribal knowledge" is the polite term for everything an organization knows that is not written down anywhere you can find it. Why the last vendor was dropped. What the client actually cares about. Which of the three versions of the template is the real one. Why the committee said no in 2024, so that nobody has to re-litigate it in 2026.
Every attempt to fix it runs into the same wall: the people who hold the knowledge are the busiest people in the building, and asking them to write it down is asking them to do a second job.
The mistake in most knowledge-management projects
The classic approach is to build a place for knowledge to go, then ask people to put it there. A wiki. A CRM with mandatory fields. A "lessons learned" template at the end of every project.
These fail for a predictable reason. Writing for a future reader is expensive, the payoff is uncertain and belongs to someone else, and the work that generated the knowledge already produced a document: the proposal, the report, the email, the deck. Asking for a second, structured version of it is asking for the same knowledge twice.
The result is a wiki that was accurate in the quarter it was launched.
The knowledge is already written down
Look at what actually gets produced in a normal week. Proposals. Close-out reports. Meeting notes. Feasibility questionnaires. Vendor summaries. Performance letters from clients. Spreadsheets tracking whatever the team tracks. Nearly every piece of tribal knowledge that matters is in one of these, stated in passing, by someone who was not trying to document anything.
The committee's reasoning is in the meeting notes. The vendor's weakness is in the close-out report. The client's real concern is in the feedback letter. The rule of thumb is in the slide deck's "lessons carried forward."
So the problem is not capture. It is retrieval, and specifically retrieval of facts and reasons rather than documents.
What capture without a reorg looks like
The approach that works has three parts, and none of them asks anyone to change how they work.
Read what already exists. Every file the team produces gets read once, properly, and what it says about people, organizations, projects, decisions, and numbers is pulled out and kept, with a pointer to the sentence it came from.
Decide what kinds of things matter. Not a taxonomy of every document. A short list of what this team actually tracks, and the details that make it filterable. A consultancy tracks clients, projects, and people; a site network tracks studies, sites, investigators, and sponsors. This takes an afternoon, and a good system will draft it from the files.
Keep the connections. The same person across a CV, three reports, and a roster is one record. The reason recorded in the meeting notes attaches to the decision in the portfolio log. This is the part that turns a pile of documents into something that can answer "why."
What to expect
When this works, the change is less dramatic than a new system and more useful. People stop asking "who would know?" and start asking the question itself. Newer staff catch up faster because the organization's memory is available to them, not just to the people who built it. And the four people who used to be interrupted all day get some of their day back.
The knowledge was never missing. It was unfindable. Those are different problems with different fixes.
How Desk helps
Desk captures what people already wrote down, without asking them to write it again.
- Upload files as they are. No templates, no mandatory fields, no reorganizing the shared drive first.
- Reasons, decisions, and rules of thumb buried in meeting notes and close-out reports become findable facts with a source.
- The same person, client, or project across a dozen files becomes one record, so "who would know?" turns into asking the question itself.
Thinking about how this would work for your team?
Let’s chatKeep reading
"Have we done this before?" is the most expensive question in your company
It gets asked a dozen times a week, it is almost always answerable from files you already have, and the answer usually arrives too late to matter.
Read more →Citations or nothing: why every answer should point to a file
An AI answer you cannot check is a confident rumor. The test of a knowledge tool is not how fluent its answers are but how fast you can verify one.
Read more →