Articles, search and the index underneath
Every team has this problem and almost none of them have a documentation problem. The answers were written; they were written in the wrong place, once, to one person.
Retrieval
Search here matches meaning rather than words, so “can I send it back after a month” finds the refunds policy even though the policy never uses the word month. What it returns is the article — not a generated summary that might be right.
Writing it
So the work here is in the shortest path from an answer someone just typed to an article that exists.
Somebody answers the refund question in Piskes Chat for the fifth time. One action turns that reply into a draft article, with the question as the title and the answer as the body, waiting for a two-minute edit. That is the whole loop, and it is the only one that has ever worked.
Policy, how-to, troubleshooting, definition, checklist, decision record, FAQ. A shape to fall into beats a blank page.
Every article carries when it was last confirmed true, and stale ones are listed for review rather than quietly rotting.
A suggestion goes to the owner, so being wrong in public is not the cost of fixing a typo.
The knowledge base itself, internal or public, with the access rules applied per reader.
Piskes Chat proposes articles as your agent types, so the answer sent is the answer you already wrote.
Piskes Agents ground on the same index, which is why they can say something true about your business rather than something plausible.
All of them promoted from chat threads. None of them written from scratch, because nobody writes documentation from scratch.
Somebody finds an answer without asking. This is the week people start believing in it.
Support replies get faster because the article is offered rather than remembered.
An article passes its review date and is listed. This is the mechanism that decides whether you have a knowledge base in a year.
Pricing
Charging per reader on a knowledge base is charging you to let people find the answer, which defeats the purpose you bought it for.
For a first internal handbook.
$0
Most teams
For a team that answers the same questions.
$19 per month
For several products or several audiences.
$59 per month
Prices in USD, billed monthly. A public space costs the same as a private one.
Questions
A wiki has no retrieval and no review dates, so it becomes an archive of things that used to be true. The two mechanisms on this page — promote from chat, flag when stale — are the entire difference.
A space can be public, on your own domain on Business. Access is per space, and an internal article never surfaces in a public search.
The index is rebuilt for that article on save. There is no nightly job to wait for and no state where the search result and the article disagree.
Markdown and HTML, keeping the folder structure as spaces. Confluence and Notion exports both land in that shape.
Not with a structure, not with a template library, and not with a documentation plan. One question, one article, and the loop that turns the sixth answer into the seventh reader’s search result.