Within the Fraunhofer initiative Mat-o-Lab we are developing a knowledge platform based on OpenSemanticLab in which shared resources such as lab space, equipment, materials and projects are managed (see Lab Inventory Management System - LIMS). Moreover, the actual project work can be documented in the same system via meeting minutes, Kanban boards and electronic lab notebook entries (ELN). Beside conventional data input and output or searching via the web interface or Python API, the focus is increasingly on delegating time-consuming activities to LLMs. In addition to retrieval augmented generation (RAG) using the phrased RDF graph and conventional document embedding, we also utilize deterministic agents. Orchestrated by the (main) LLM, these agents perform specific tasks such as completing a form or creating an evaluation diagram via typed and data plot. In this way, we hope to generate increasingly large amounts of structured, validated and AI-accessible data for the organizational and scientific processing of projects without additional effort and technical hurdles for users.