An internal AI knowledge base built from your company documents
Policies, product documentation, contracts and past project materials live in several places today: SharePoint, file servers, inboxes. The internal AI knowledge base answers your colleagues' questions from these documents, and every answer links back to the original. We build it as a RAG system: before answering, the model receives passages retrieved from your own documents, with no retraining. Pilot in 2-3 weeks, implementation in 6-12 weeks; delivered by a Budapest-based team.
The next step is the AI opportunity assessment: from HUF 500,000 + VAT.
Updated:
A knowledge base spreads when people can trust the answers. That is why every answer carries a source: it shows which part of which document it rests on, and the document opens with one click. Where there is no reliable source, the system does not invent an answer; it reports that the knowledge base holds no material on the question.
At an insurance company, the knowledge base built this way over 12,000 documents was in use by 78% of staff within the first month. The retrieval layer works with semantic chunking and source citation, and the documents sync through the SharePoint API. Across our 46+ completed projects, automated processes reach a 40-45% efficiency gain; typical ROI is 6-12 months.
Access rights hold inside the knowledge base too: a colleague gets answers only from what they can see in the source system. A full team carries the work: a project manager, senior developers, and an AI DevOps engineer. We run the EU AI Act risk classification at the start of every project; an internal knowledge base typically falls into the minimal or limited risk category.
What the internal AI knowledge base does
Answers from your company documents, with sources
A colleague asks; the system answers from policies, product documentation and past materials. Next to every answer sits the source: which document, which section.
Access control that follows the source system
The system follows the permissions of SharePoint and the file server: everyone gets answers from what they can access today. Confidential material never enters an answer the asker could not see.
Automatic updates
A new policy, a new product description, a changed contract template: the document is updated in the source system and the knowledge base follows. There is no second copy to maintain.
Faster onboarding, fewer interruptions
New hires ask the knowledge base and get an answer with a source; experienced colleagues stop losing their time to the same repeated questions.
How does the implementation work?
- 1
AI opportunity assessment (from HUF 500,000 + VAT)
We review where the documents are, in which formats, who may access what, and which questions the system has to answer. We tell you what is missing from the knowledge base. The output is a written plan with cost and payback estimates.
- 2
Pilot
We run the system live on one bounded set of documents with one team, and measure answer accuracy, the correctness of the source citations and the usage rate. A small proof of concept is ready in 2-3 weeks.
- 3
Implementation and operations
Integration with SharePoint, the file server and your internal systems, access control, training. Most projects go from kickoff to production in 6-12 weeks. Source code, models and documents remain yours; we also offer monthly managed operations.
Why Leventech
We build and operate production RAG systems
Retrieval pipelines, prompt-injection defence, per-component failure handling: the internal knowledge base we built for an insurance company was used by 78% of staff in the first month.
Our team has built production systems since 2013
46+ completed projects, 96% client satisfaction. What we build, we also operate and maintain.
EU AI Act and GDPR built into development
Risk classification at project start, access control and logging in the design. We have experience implementing high-risk AI systems as well.
Common questions
In retrieval-augmented generation, the language model receives relevant passages retrieved from the organization's own documents before answering, and grounds its response in them. The approach ties answers to the company's knowledge base without retraining the model and reduces hallucination. Fine-tuning comes into play when the domain language justifies it.
An AI assistant or integration project is usually €15-40K. The entry point is the AI opportunity assessment: from HUF 500,000 + VAT; it shows which set of documents to start with. After the assessment you get an exact quote with no hidden costs.
Every answer carries a source: it shows which part of which document it rests on, and the source opens with one click. In the pilot, your own experts check answer accuracy against a measured score.
It answers from them only for people who have access in the source system. We take permissions from SharePoint or the file server and check them on every question.
Not necessarily. When needed, we deploy on-premise or on your private cloud, with open-source models running on your servers; the documents never leave your environment.
The internal knowledge base answers your colleagues from the company's full document base, according to their permissions. The customer service AI answers customers' email and chat questions from a knowledge base designated for that purpose. The two systems can share the same retrieval layer.
Still have questions?
Start with a conversation
Tell us where your documents live today and what your colleagues ask most often; we tell you where to start.
What would you automate in your business?
Tell us briefly about the task. We will reply by email to arrange an initial conversation.
The next step is the AI opportunity assessment: from HUF 500,000 + VAT.
What happens after you send?
In the first conversation we review the task, your existing systems and the outcome you need. Then we discuss whether a detailed assessment would help. We provide a proposal before any paid work.
Or reach us directly:
[email protected]
Monday-Friday, 9:00-17:00 CET