Every company has tried a wiki. Most have several, abandoned in sequence, each with a burst of initial content and then silence.
- ›Why wikis decay
- ›What changed with retrieval
- ›What is worth documenting
- ›Keeping it honest
Why wikis decay
Writing is unrewarded. The cost is paid by the author, the benefit accrues to a stranger later. That asymmetry is structural and no amount of encouragement fixes it.
Finding fails. People search, get poor results, and stop searching. Once the habit breaks it does not return, and the content becomes invisible regardless of quality.
Nothing signals staleness. A page written two years ago looks identical to one written yesterday, so readers cannot tell what to trust.
No ownership. Content nobody owns becomes content nobody updates.
What changed with retrieval
The finding problem is now genuinely solved. Asking a question in natural language and getting an answer drawn from your own material, with citations, works.
That shifts the bottleneck from finding to capturing, which is a better problem to have, and it changes what is worth writing: you no longer need a navigable structure, you need accurate content that retrieval can reach.
What is worth documenting
Be selective. Comprehensive documentation projects fail; targeted capture works.
Decisions and their reasoning. The highest-value category and the one nobody writes down. Not what was decided, but why, what alternatives were considered, and what was expected.
Outcomes. What actually happened afterwards. Decisions without outcomes cannot teach.
The things people say "you have to know" about. Client quirks, system gotchas, the reason a process is unusual.
Answers to repeated questions. If it is asked monthly, write it once.
How to do the things only one person knows. The bus-factor list.
What to skip
Meeting minutes nobody reads. Aspirational process documents describing how work should happen. Anything already recorded structurally in your systems. Org charts that go stale in a month.
Tip: Capture at the moment of decision, not in a documentation sprint. Two sentences written when a decision is made beats a page written six months later from memory, and it is far more likely to happen.
Keeping it honest
Owners per area. Not a central documentation team below significant scale; the person closest to the work.
Freshness expectations. Mark content with how long it should be trusted, so stale pages announce themselves.
Treat wrong answers as bugs. When someone gets a wrong answer, fix the source rather than correcting the person. The correction should persist.
Prune. Deleting outdated content is a contribution.
FAQ
Should we document processes or outcomes?
Both, differently. Processes for things that must be consistent; decisions and outcomes for everything that involves judgment.
How do we get people to write?
Lower the cost: two sentences at decision time rather than a document later. And make the benefit visible by showing people getting answers from what was captured.
What about knowledge in chat?
It is there and effectively unretrievable. Extract decisions into durable form and treat the chat log as an archive. See the company brain.
Brainis captures decisions, institutional memory, and documents, retrievable by Cortex with citations across every module. See Cortex.
Sharing insights on business operations, AI, and modern team management.
Run your company on Brainis
All 11 Operating Systems on every plan, from $29 a month. No per-seat pricing — you pay for AI capacity, not headcount.
See pricing