Your own knowledge,searchable
Most businesses know more than they can find back. Manuals, protocols, quotes, email threads: it is all there, spread across folders nobody fully knows. A knowledge base assistant makes that searchable in plain language, with the source attached so you can check the answer.
Does this sound familiar?
- 01New staff ask the same questions an experienced colleague answers from memory.
- 02The knowledge lives in people and in folders, and leaves when those people leave.
- 03Searching by file name only works if you already know which document you are looking for.
- 04An ordinary chatbot makes up a plausible answer instead of saying it does not know.
What you get.
- Your documents opened up in a searchable layer (RAG), so questions can be asked in plain language
- Answers with source references, so you can check where they come from
- An explicit 'that is not in here' instead of a made-up answer
- Access per role, so not everyone can reach everything
- Available where your people already work: in the browser, by email or in an existing screen
Operating it yourself says more than a description.
Pick a question and watch the assistant answer from a knowledge base, citing its sources, and honestly flag when the answer is not in there. This is a demo on sample data, not client work.
Demo · prepared in advance · illustrative · no live model
Choose a question about Aether AI
Illustrative demonstration with examples prepared in advance. No live AI call and no API key needed; the answers are fixed and contain no invented figures or names.
Four steps, in this order.
We decide the source
Which documents are authoritative, and which are out of date? This is the real work, and it is work you know and I do not.
I build the layer
Your documents are made searchable, while keeping track of where each thing comes from.
We test on real questions
A set of questions you know the answer to, to measure whether it gets them right.
Extend per department
Once it works for one group, and only then, we extend it.
Lead time: one to four weeks, depending on the state of your documents.
PostgreSQL · pgvector · Supabase · Python · Anthropic · Open modellen
What people ask me about this.
Q1Does my company knowledge leave the building?
Only if you want it to. This technique can run entirely on your own server, with an open model, so that no document goes to an external party. If you do opt for a cloud model, we set out in a data processing agreement what is sent and what happens to it.
Q2What if the answer is wrong?
That is why the source is always attached: you can trace every answer back to the document it comes from. And if the knowledge base does not contain the answer, the system should say so instead of making up something plausible. That is exactly the difference the demo next to this shows.
Q3Do my documents need to be in order first?
Not perfect, but honest. If three versions of the same protocol are floating around, someone has to decide which one is authoritative. That judgement is yours; what I can do is make visible where the contradictions are.
Q4Can what someone sees differ per department?
Yes. Access per role is part of the setup, and with sensitive data it is not an extra but a condition.