The impact counter
The article will show how often it was suggested and opened — and how often no ticket came of it. The number every knowledge base has been missing: is the writing worth it?
The knowledge base
that shows up in the ticket.
Documentation that comes to where the question is asked — instead of sitting where nobody looks.
Page tree + Page
SalesApp public
Adding a second delivery address
Quotes & discounts
Interfaces public
Supplier portal: formats
Parts list import 🔒
Operations 🔒 internal
3 subpages inherit “internal”
Etappe Book · SalesApp
Public — the customer sees this page Adding a second delivery address
A second delivery address is stored on the customer, not on the order. That way it is available in every future ticket.
Open the customer → “Addresses” tab
Choose “+ Address” and mark it as a delivery address
In the quote, the choice appears automatically in the header
Last surfaced in #1802 and 4 other tickets found by meaning, not by keyword
The cycle
A wiki is a place you go to. Book comes to you.
An article doesn't wait to be searched for — it shows up where the question is asked. And knowledge doesn't end: it is written, found, put to work, goes stale and gets maintained. That's why this is a circle.
✦ what the AI contributes
- 01 Write Blocks, images, structure — no markup ✦ Outside agents may create and maintain pages.
- 02 Classify Areas, page tree, public or internal Visibility is a decision, not a suggestion.
- 03 Find Full-text search per area ✦ Answers the question directly, sources cited.
- 04 Work The article appears in the ticket before anyone asks ✦ Picks it by meaning, not by keyword.
- 05 Measure PLANNED How often has this article helped?
- 06 Maintain Changes visible, history traceable ✦ Summarises every edit in one sentence.
- ↺ Knowledge doesn't end — it comes back.
01 · Write
Structure instead of formatting work.
A block editor, not a word processor. Headings, lists, tables, code, images — every element is a block you can grab and move. No Markdown to learn, no HTML to repair.
Every area has its own page tree, as deep as you like, with icons per page. A wastebasket instead of deleting: restorable for 30 days. And an archive for everything that no longer applies but explains why things are the way they are.
02 · Classify
Public or internal. Inherited.
Every page is public or internal — a decision, not a suggestion. It inherits down to every subpage: the new page under an internal chapter is internal, without anyone having to remember.
And the separation doesn't stop at the system boundary: when Flow suggests an article on a ticket, the customer is never even sent internal docs — not hidden, not there.
03 · Find
Search. Or just ask.
Full-text search across titles and content, per area — fast, no configuration.
And whoever asks a question in full sentences gets an answer instead of a results list — sources cited. The AI first picks the relevant pages, then answers from exactly those. That detour is the guarantee that the answer comes from your pages, not from model knowledge — and that the citations are right.
How do I add a second delivery address?
That's in Etappe Book — here's the guide:
Adding a second delivery address
Etappe Book · SalesApp
Answered from Etappe Book — no ticket created
04 · Work
The article comes to the problem.
When someone opens a ticket in Flow, Etappe checks whether Book has something on it — by meaning, not by keyword. Matching articles appear as cards right on the ticket. The customer only sees the public share; can be switched off per customer.
Questions that need explaining answer themselves — before someone on your team types them for the fifth time. The best ticket is the one that never happens. That's exactly what Book is for.
SCREENSHOT TO FOLLOW Ticket with the “Helpful articles” section — two cards, area as badge
06 · Maintain
Knowledge goes stale when nobody looks.
Book remembers the state before every change and has the AI summarise in one sentence what changed — not “page updated”, but what is actually different. Changes go to Slack, bundled per page: one message per session, not one per keystroke.
Plus a complete activity log and an author page per person — showing who carries the knowledge, and where one person is a single point of failure.
SCREENSHOT TO FOLLOW Change history of a page with the AI summaries per edit
Writable — for humans and for machines.
A token created from the interface and bound to a person gives an outside AI agent access to the pages API: list, read, create, update. Your coding agent documents what it built — in the same run it builds it. No “docs will follow”.
- Short-lived. The token lasts about an hour; a new one replaces the old.
- Bound to a person. Changes carry that person's name in the history — traceable who let an agent write.
- Same rules. Visibility and inheritance apply to agents just as to humans.
Flow lets outside agents work the ticket. Book lets them write the docs. Nobody else offers this pair.
Moving in
Confluence out. Book in.
Existing documentation comes along — structure, images and links survive. No copy-paste project, no lost hierarchy, no dead references.
The import is repeatable: a single page can be imported again without touching the rest. Proven with real content — two areas, the complete tree, including the macro content that standard exports choke on.
SCREENSHOT TO FOLLOW Before/after: the Confluence tree next to the Book tree
Evidence, not claims
- Runs in production
- In daily use as the internal knowledge base of the agency that builds it
- Migration proven
- Two Confluence areas imported with their complete structure
- AI with a cap
- Daily limit for doc questions; without a key the feature simply stays off
- Separate environments
- Test and production have their own data — no accidental overwriting
PLANNED
What comes next
Decided and scheduled, not yet shipped — which is why it stands here as an outlook, not in the present tense.
Document from the ticket
A ticket is solved, the answer sits in the thread — Etappe will propose an article draft from it. The moment knowledge is created is the moment it must be documented.
One search across both
⌘K will find tickets and articles in the same list. Whoever searches is looking for an answer — not a system.
Ask the docs and the history
The question goes to Book and to the ticket archive: “How did we solve this for customer Y back then?” The answer sits in both.
These docs run on Book themselves
docs.etappe.ai is not a demo but a live Etappe Book instance — the same reading, searching and linking you get for your own documentation. It is written in German for now.
Frequent questions
Do we need Flow to use Book?
Book comes with Flow — one price model. It is built as the second half of the same system, not as a standalone wiki.
Do customers see our internal docs?
No. Every page is public or internal, and the setting inherits down to every subpage. In the ticket, the customer is only ever delivered the public share.
Can we bring our Confluence content?
Yes — with structure, images and links. The import is repeatable and recognises what has already been moved.
Does the AI answer from our docs or from the internet?
From your docs. It first picks the relevant pages and formulates the answer from them — sources cited under every answer.
Can we switch the AI off?
Yes. Without a stored key the AI features stay off; writing, searching and reading are untouched.
Can an AI agent document for us?
Yes. Via a short-lived token bound to a person it may read, create and update pages — under the same visibility rules, traceable in the history.
Book only comes with Flow —
two modules, one price.
Sign up, short onboarding, straight to work — free plan, no credit card. Just bring your Confluence.