Development of an AI System for Information Retrieval in Corporate Wiki via Mattermost : An AI System for Information Retrieval in Internal Documentation

Artificial intelligence (AI) is increasingly used today to help companies manage and find information in large volumes of documents. One rapidly growing area is so-called Retrieval-Augmented Generation (RAG) systems, where the AI model can search the company’s own documents to provide more accurate and up-to-date answers to user questions. A common problem with such systems is that they often require information to be sent to cloud services, which can pose a risk for companies that want to keep their information secure. Additionally, updating the document that the system uses can be difficult, and using AI on acompany’s own servers can be technically challenging. In this thesis, we have developed an AI system based on RAG that can run locally on the company’s own computer without the need to send information to the cloud. The RAG-system uses a technique where the company’s internal documents are converted into a format that the AI can easily search, and when the user asks a question, the AI helps the system to find the relevant parts of the documents to formulate an answer. We have tested the system using documents from a company’s internal wiki and found that it works well for providing accurate answers, even to more complex questions that require the AI to connect information from multiple parts of the document. The system runs effectively on local hardware with a dedicated graphic card without the need to connect to cloud services. The results show that the system has great potential, with opportunities for further improvements and expansion.