This paper proposes a reliable way to compensate for the limitations of the current generation of large language models in relation to common business use cases by adapting an automated chat agent with the ability to use self-implemented tools. Using the aforementioned approach, context size limitations were extended using a semantic search database and conversation abilities over a proprietary dataset (as well as a few complementary behavioral patterns) with support for mathematics introduced. The paper clarifies the data extraction and processing methods and the architecture used for agent construction. Lastly, external and implementation-related limitations are mentioned, along with an evaluation of the obtained results.