Your new colleague asks how a handover works. You look for the most recent suitable prompt file, then a chat and finally an older set of instructions. The problem is not too little text. What is missing is a recognisable place where the team’s valid way of working is written down.
BusinessCloud has “Knowledge” for this: knowledge bases made up of pages and subpages for your own company. Technically, this route is based on Nextcloud Collectives. It is a structured collection of knowledge, not an automatic AI document search. This difference makes it clearer what it is suited to.
A page has a different purpose from a prompt attachment
A prompt attachment supplies material for one specific AI task. A knowledge page, by contrast, describes a state of affairs that people can find again, read and maintain. The same set of instructions can support both, but its existence alone does not give an AI access to it.
Our own example is a small workshop team. It needs the procedures for the start of a shift, issuing materials and handover. Instead of attaching a long text with every special case, it creates a knowledge base called “Workshop handbook” with a short start page. The three procedures become separate subpages.
Build the hierarchy around the real work
The start page explains what the handbook applies to and which page helps on which occasion. “Start of shift” describes the first check, “Issuing materials” the confirmed rules for parts that are needed, and “Handover” the state of work that has to be passed on. A page should answer a recognisable question.
Workshop handbook
Start of shift
Required documents
Pending approvals
Issuing materials
Handover
Completed
Blockers and open decisionsThis tree, which we designed ourselves, is an idea for content, not an automatically created knowledge base. It shows why subpages help: new people do not have to filter the relevant way of working out of a long catch-all document. A special rule can sit with the procedure it affects instead of being repeated in slightly different forms in various places.
The knowledge base belongs to a team
In KENJTEC’s specific implementation, a knowledge base belongs to a team. When you create one, you choose people or teams from your company; the search is limited to this company context. To achieve this, KENJTEC adds its own controls and configuration around the Nextcloud foundation. General upstream sharing options are therefore not a blanket promise for this offering.
The KENJTEC documentation on “Knowledge” describes the specific route via “New collective” and pages with subpages. Plan membership deliberately according to what the work requires. An internal knowledge base here is not a public help centre open to any customer.
Ownership matters too: the storage used by a knowledge base counts towards the person who owns the team. A knowledge base is deleted in “Knowledge” itself and first goes to the bin. The last member cannot simply leave a team that has a knowledge base. The structure is therefore part of the working assets, not just a loose list of links.
Keeping things current needs a specific person responsible
On every important procedure page, write down its scope and its confirmed status. If the rule for releasing materials changes, you edit exactly the rule concerned. Add briefly which decision replaces the previous state. A responsible person within the team can check that the content is up to date; that is an organisational arrangement made by the team, not an automatic product role.
Avoid having two versions that are both labelled as binding at the same time. An old file can be kept as background while the page clearly points to the valid procedure. The gain is a recognisable order of precedence. This does not mean that simultaneous joint editing is promised for BusinessCloud.
Maintaining knowledge is not yet a RAG pipeline
If you want to work on a section externally with AI, you choose it and hand it over yourself, deliberately. This does not make the AI search all company documents automatically. There is no evidence of automatic indexing, OCR or a general file RAG function for this route.
The technical description follows the current code and the feature list; a full test with two real companies was not carried out here. For a start of your own, a small knowledge base with three pages and one traceable change is enough. That will show whether the next person finds the valid procedure without having to know the original prompt or chat.
Sources and code review status: 25 October 2026. The general Collectives foundation and KENJTEC’s specific company implementation were considered separately.
