The problem it solves
Every organisation over about thirty people has knowledge scattered across a document store, a wiki, an email archive and several people's heads. New staff ask questions that are answered somewhere; experienced staff spend their time answering them.
The application lets someone ask in plain language and get an answer with a citation, so they can check it. The citation is what makes it trustworthy enough to use.
Document quality decides the outcome
Every internal knowledge project turns into a document housekeeping project by week three. We treat that as part of the work rather than as a surprise, and it is usually the most valuable part.
- One authoritative version per policy, older ones archived out of the index
- Dates and owners visible, so currency can be judged
- Scanned documents processed to text, or they are invisible
- Drafts, personal folders and superseded material explicitly excluded
Permissions are enforced at retrieval
The fastest way to create a serious incident is to index HR files and salary data into a system everyone can query. Access control filters by the asking user's permissions before anything reaches a model.
This is designed in from the start. Retrofitting per-user filtering onto a system already in use is unpleasant and, in the meantime, risky.
Put it where people already are
The chat tool the team lives in, the intranet, or the helpdesk. A separate destination gets visited enthusiastically for three weeks and then forgotten.
This single decision affects adoption more than any accuracy improvement.
Measure with real questions
Collect fifty real questions from staff with agreed correct answers and use them as the evaluation set, rerun on every change to chunking, retrieval or prompting.
Also track the questions it could not answer. That list is the most useful documentation backlog most organisations have ever had.