Knowledge base preparation
We organise the sources – documents, web pages, databases and systems – and agree how often they should be refreshed.
AI services
AI chatbot on your company knowledge
We implement an AI chatbot on your company knowledge that answers employee and customer questions based on your documents, procedures and data. RAG and semantic search keep the answers up to date, grounded in sources and aligned with each user’s permissions.
What problem a company knowledge chatbot solves
Company knowledge is scattered: procedures in documents, answers in old emails, product specifications in PDFs and agreements in the heads of experienced employees. New starters take a long time to get up to speed, and teams waste time searching for information and answering the same questions over and over.
General-purpose AI tools do not know your documents and can give convincing but wrong answers. A chatbot built with RAG (Retrieval-Augmented Generation) first finds the right passages in your company knowledge base and only then formulates an answer – and it shows the sources it relied on.
This works well as an internal assistant for employees, as support for customer service teams, and as an intelligent search engine for product catalogues or technical documentation.
Where a RAG chatbot delivers the most value
- Answering employee questions about procedures and policies.
- Supporting customer service and technical teams.
- Faster onboarding of new employees.
- Searching documentation and product catalogues.
- Answers with sources that are easy to verify.
RAG implementation – what the service covers
We build a complete solution: from preparing the knowledge base, through semantic search, to the interface and integrations.
Semantic search
We index content so that search understands the meaning of questions, not just keywords. We use PostgreSQL and pgvector.
Answers with sources
The chatbot shows the documents its answer is based on, and when it does not know the answer it says so instead of guessing.
Knowledge permissions
Users only receive answers from documents they have access to in your company systems.
Interface and channels
The chatbot can live in a web app, intranet, website or the tools your team already uses.
Quality evaluation
We test answers against a set of control questions and monitor quality and costs after launch.
How we implement a RAG chatbot
The key to an effective chatbot is well-prepared knowledge and thorough testing on the questions real users ask.
Consultation and goals
We agree who will use the chatbot, what questions it should answer and from which sources.
Knowledge inventory
We collect and organise documents and data and agree update and permission rules.
Prototype and test questions
We build a prototype and check answer quality against a question set prepared with your team.
Rollout
We launch the chatbot in the chosen channel, with logging and a way for users to rate answers.
Growth
We fill knowledge gaps based on unanswered questions and add further sources.
Technologies
RAG and semantic search technologies
We store the knowledge base in PostgreSQL with the pgvector extension and use Redis for caching. We work with OpenAI and Anthropic models, build the backend in Python and the interface in React or Angular. We monitor answer quality and costs in Langfuse.
- OpenAI
- Anthropic
- PostgreSQL + pgvector
- Redis
- Python
- React
- Angular
- Langfuse
- Model Context Protocol
- Docker
Example use case
An AI assistant for a customer service team
Customer service agents at a company serving business clients search hundreds of pages of terms, manuals and specifications for answers. New employees often have to ask more experienced colleagues.
The company knowledge chatbot answers agents’ questions by quoting the relevant passage of a document with a link to the source. Questions it could not answer go to a list of knowledge base gaps, so the chatbot becomes more useful over time.
An illustrative scenario showing the typical course and scope of such a project.
Frequently asked questions
What is RAG?
RAG (Retrieval-Augmented Generation) is a technique in which an AI model first searches your company knowledge base for the right information before answering. Answers are therefore based on up-to-date documents and can cite their sources.
Will the chatbot make up answers?
We design the chatbot to answer only from the documents it finds and to say when it does not know. We verify quality against a set of test questions.
Does the chatbot respect document permissions?
Yes. Users only receive answers from documents they have access to – permissions can be taken from your company systems.
How is the chatbot’s knowledge kept up to date?
Sources are re-indexed automatically on a set schedule or when a document changes, so the chatbot uses current information.
Can the chatbot run on our website?
Yes. The chatbot can work internally for employees or publicly for customers – for example on your website or online store.
Experience since 2009, now powered by AI
Since 2009
17 years of building enterprise-class web applications, custom business systems and mobile apps.
Industry know-how
Many projects for automotive, logistics, healthcare, food service and catering, construction and finance.
From idea to maintenance
Consulting, design, development, integrations, rollout, training and ongoing growth in one place.
Data security
Solutions designed in line with GDPR and the EU AI Act, with access control and transparent costs.