How to keep your internal knowledge base alive
If you’re an EM, your job isn’t to write every doc. Your job is to keep the system trusted. Most internal knowledge bases die the same way: the content gets stale, search fails, and new hires stop looking.
You don’t need a new tool. You need a few habits that keep the content current and owned.
1) Attach knowledge to real incidents
If a runbook or answer was used during an incident, capture it within 24 hours. That’s when context is fresh and the team will actually do it.
Add:
- the symptom in plain language
- the fix that worked
- the dashboard or query link
- the owner who can confirm updates
2) Give every entry an owner
A knowledge base without ownership is a pile of anonymous notes. Every entry should have a responsible team or person.
Ownership doesn’t mean they wrote it. It means they’re the person to page when it becomes wrong.
3) Kill or merge duplicates
When you see two entries that say the same thing, merge them. Duplicates are worse than missing content because they create doubt.
Make a rule: one answer per question.
4) Review on a schedule
Pick a cadence that matches risk:
- critical runbooks: monthly
- common questions: quarterly
- low-risk how-tos: twice a year
Add a short “last reviewed” note to keep it honest.
5) Add update triggers
Update an entry when:
- the owner changes
- a dashboard or query link changes
- the workaround becomes a real fix
5) Use tags that mirror how people think
People search for what they experience, not what your org chart says. Use tags like:
- latency
- deploy rollback
- payments
- data pipeline
Avoid internal project names unless they’re common in Slack.
6) Turn gaps into tasks
When someone asks a question you can’t answer quickly, that’s a task. Record it immediately and assign an owner. That’s how the knowledge base grows in the right direction.
How we do this at DawnOps
We tie knowledge updates to real work:
- after incidents, we capture the single most repeated question
- we assign an owner and a review date when the entry is created
- we link answers to the exact dashboards and runbooks used in the moment
The small win
If you do nothing else, do this: after every incident, write one clean entry that answers the single most repeated question from that incident.
That’s enough to keep the knowledge base alive.