Four pages called Onboarding come back as one sentence: this one is current, these two are superseded, this fourth is a different team. Then the runbook gets rewritten from the sources.
The deployment runbook was last edited fourteen months ago by someone who left in March. Search returns 60 results and four of them are called “Onboarding”, written in 2023, 2024, and twice in the same quarter by two teams who did not know about each other.
A companion reads all 60 candidates with their version metadata and reports what each one claims, when it last changed, and where two of them contradict each other.
01
Which of these 60 pages is the one that is true?
A companion searches, reads every candidate with its version metadata, and answers with evidence rather than a ranking. Four pages called “Onboarding” become one sentence: this one is current, these two are superseded, this fourth is about a different team.
The contradictions are the output worth having: the two pages that disagree, named side by side, with what each one claims. Sorting that out is operations work rather than search.
02
Pages written from the source, not from memory
Updating the runbook means reading the pipeline config, the incident thread and the deploy dashboard, then writing prose. A companion does that reading in the tabs you are signed into and drafts the page: the current steps in order, what changed since the last version, and the parts it could not verify flagged.
New pages are created under a parent so they land in the hierarchy rather than at the root, and everything is yours to approve before the space sees it.
03
How does an edit avoid overwriting somebody else?
An update carries the page’s current version number, which is Confluence’s own guard against two people overwriting each other. That means an edit is always made against the version the companion actually read.
Footer comments are readable and writable too, so the question a page raises can be answered on the page rather than in a channel nobody searches.
04
Write the page from the dashboard and the ticket
The dashboard, the pipeline config, the ticket, the vendor doc, the email where the decision was actually made. A companion reads those signed in as you and then writes the page. Turning scattered sources into one document is the research motion with the wiki as the destination.
Atlassian Intelligence, as of this writing, answers questions about what a space already says, licensed with the product. Writing the page that does not exist yet is the pass above.
Yes. Confluence is a native integration. A companion lists spaces and pages, searches with CQL or plain text, reads a page with its body and version, walks child pages, creates and updates pages, adds and reads footer comments, and deletes a page to trash. Publishing waits for your approval.
That is the main use. A companion reads the dashboards, tickets and threads you are signed into, drafts the page with the parts it could not verify flagged, and creates it under the parent you choose once you approve the draft.
Ask for the contradictions rather than the list. A companion searches, reads the candidates with their version metadata, and reports which page is current, which are superseded, and where two pages disagree.
Edits and new pages are staged for approval, so nothing appears in a space your team reads without a person saying yes. Updates carry the page’s current version number, which is Confluence’s own protection against overwriting somebody else’s change.
Both are native integrations, so one companion reads the tickets in a release and writes the release page that describes them. Download Strawberry free, paid plans from $20 a month, and connect the two together.
Confluence Strawberry