How I run documentation
This is the operating model I use to plan, produce and maintain documentation for a team that supports many applications and processes. It's how I'd set up documentation as a lead.
flowchart LR
A[Intake] --> B[Prioritise] --> C[Plan & scope] --> D[Write] --> E[Review] --> F[Publish] --> G[Maintain]
G -. feedback & gaps .-> A
1. Intake
- One front door for requests: a Jira ticket with a short template – audience, purpose, deadline, SME and related tickets
- Requests from release planning are added automatically for anything user-facing
2. Prioritise
I rank requests on three questions:
| Question |
Higher priority when… |
| Impact |
Many users rely on it, or it supports a release, audit or UAT |
| Risk |
Missing or wrong content could cause errors, rework or compliance issues |
| Effort |
It can be delivered in one cycle with available SMEs |
3. Plan and scope
- Agree the audience and the questions the document must answer
- Choose the right content type – process overview, SOP, reference, release note or FAQ
- Book SME time early; a document without an SME is a guess
4. Write
- Start from a template for the content type, so pages look and read the same
- Follow a style guide: plain language, active voice, task-based headings, consistent terms (see the glossary)
- Write for people and AI search: one topic per page, descriptive headings, no key facts hidden in images
5. Review
| Review |
Who |
Checks |
| Technical |
SME or developer |
Accuracy and completeness |
| Peer / editorial |
Another writer or me |
Clarity, structure, style |
| Audience |
A target user |
Can they complete the task without help? |
6. Publish
- Publish in the agreed space, add labels and link the page from its parent process
- Close the Jira ticket with a link to the page
- Announce major changes in release notes or the team channel
7. Maintain
- Every page has an owner and a last reviewed date
- Quarterly review of high-traffic and process pages
- Retire or archive outdated pages rather than leaving them to confuse readers – and AI agents
How I measure documentation
| Metric |
What it tells you |
| Pages published and updated per cycle |
Throughput |
| Review turnaround |
Where work gets stuck |
| Page views and search terms with no results |
What people need and can't find |
| Repeat questions to SMEs or support |
Gaps the documentation should close |
| AI agent answers without a source |
Missing or unclear content |