AI knowledge base for companies, with a source on every answer
RagINS is an AI knowledge base for companies: employees ask questions in Arabic or English and get answers drawn only from the company’s approved documents, each with a citation to the exact passage. It is built for HR, IT, operations, procurement and legal teams in Egypt, Saudi Arabia, the UAE and the wider GCC.
How the AI knowledge base for companies works
- 01
Add your documents
Upload PDFs, Word (.docx) files and text files such as TXT, Markdown and CSV, several at a time, or paste text; documents, wikis and drives are indexed with their permissions preserved. Duplicates are detected, and each file is indexed by an ingestion job you can follow.
- 02
Organise and restrict
Group documents into collections and set who may read each one — by role, department or named user — with a sensitivity label from Public to Restricted.
- 03
Ask in plain language
Employees ask in Arabic or English. Search combines exact keywords with meaning, then reranks the passages and removes duplicates.
- 04
Get a cited answer
Each grounded answer cites the page and section it came from. Open the passage with your search terms highlighted.
- 05
“Not enough evidence” instead of a guess
When the documents do not hold the answer, RagINS says so rather than inventing one.
- 06
See what people ask
Analytics show the questions asked and each answer’s verdict, and one-click CSV reports cover top questions, activity, access review and an audit extract.
For companies in Egypt, Saudi Arabia, the UAE and the GCC
Arabic and English documents together
Arabic text is normalised — diacritics, tatweel, and alef and ya forms — so spelling variants still match. With a multilingual embedding model, an Arabic question finds an English policy, and the reverse.
Saudi Arabia and the UAE
Where documents and model calls are processed is agreed in the deployment: an agreed cloud region, your own data centre, or hybrid.
Egypt
The whole interface reads in Arabic, right to left, for employees who prefer it, and Epsilon’s Egypt team runs the pilot with you.
Groups across the GCC
Each company is its own isolated workspace, with its own users, documents, collections and audit trail.
How to choose an AI knowledge base for your company
- 1
Citations you can open
Every answer should point to the page or section it came from, and you should be able to read that passage.
- 2
Honest refusals
Ask a question your documents do not answer. A good knowledge base says so.
- 3
Permissions before search
Access rules should be applied before results are ranked, not filtered afterwards.
- 4
Arabic on your own documents
Test spelling variants and Arabic-to-English questions on your real files.
- 5
Where data and model calls go
Ask which components run where, and whether any outside AI service sees your documents.
- 6
Administration and audit
Look for two-factor sign-in, directory provisioning (SCIM), retention rules and a tamper-evident audit trail.
- 7
How content gets in
Upload files or paste text, and index documents, wikis and drives with their permissions preserved; the sources you use are connected in the pilot.
Frequently asked questions
What is an AI knowledge base for companies?
A system that turns company documents into answers employees can ask for. RagINS finds the relevant passages, writes the answer from them only, and cites the source.
Is RagINS a chatbot for our website?
RagINS is built for your employees: they sign in to your company workspace, and every answer respects that person’s permissions. Any other channel you need is connected in the pilot.
Which file types can we add?
PDF, Word (.docx), and text files such as TXT, Markdown, CSV, TSV and JSON, or pasted text — plus documents, wikis and drives indexed with their permissions preserved. Scanned documents and any other sources you use are connected in the pilot.
Does it work in Arabic?
Yes. Questions, answers and the interface work in Arabic and English, and Arabic spelling variants are normalised. With a multilingual embedding model, an Arabic question finds an English document, and the reverse.
Who can see which documents?
Each document can be limited by role, department or named user, and those rules are applied before search. Someone who loses access to a document can no longer open its citation.
Are our documents used to train public AI models?
Not by default. Model providers are approved per company, deny-by-default, and a built-in answer mode makes no outside call at all.
Can it sync with our file shares and drives?
Yes. RagINS indexes documents, wikis and drives with their permissions preserved, so people only ever retrieve what they are already allowed to see. The sources you use are connected in the pilot.
See the AI knowledge base on your own documents
Bring a representative set of documents and the questions your teams ask. The Epsilon team shows RagINS on them and scopes a pilot. The enquiry does not create an account.
Related pages
Chat with your documents
Ask your PDFs, Word and text files questions in Arabic or English and get answers with the source page cited, inside a private, permission-aware company workspace.
Read the pageArabic enterprise search
Search and question answering over Arabic and English company documents, with Arabic normalisation, cross-language retrieval and permission-aware, cited answers.
Read the pagePrivate AI on premise
RagINS deployed inside your own infrastructure, with a built-in answer mode and self-hosted embeddings, so documents need not reach an outside AI service.
Read the pageAI document search in Saudi Arabia
Enterprise RAG for Saudi organisations: cited answers from Arabic and English documents, with on-premises or hybrid deployment and a team in Saudi Arabia.
Read the page