While companies have adopted AI and automated work processes, a key unresolved issue remains-how to test new digital solutions for process automation-both from an environmental and economic perspective, quickly and cost-effectively. Companies have not always mapped their processes, so using plain text process descriptions to get process maps is challenging. Our research aims to understand how AI could be used in the context of process extraction from company text documents to simulate reference process models in a logistics laboratory. We propose a workflow and tools for adapting AI in process extraction in a laboratory context. Extracted and classified processes are paired with reference processes model, SCOR model KPIs, and software dedicated to logistics, enabling us to build up the simulator quickly and cost-effectively. For classification, we use the EstBERT (Estonian Bidirectional Encoder Representations from Transformers) language model, which is explicitly devoted to Estonian words and fine-tuned for the definition of logistics process classes based on textual processes descriptions extracted from bachelor’s thesis and project data. The trained model showed an accuracy of 80% in classifying the processes. We validate our method by applying logistics company examples, using their process descriptions and showing how we could integrate our method into logistics laboratory work. A successful case study demonstration extends the current process mining approach by broadening the scope of valid input formats. It suggests that natural language-based representations of processes can be directly converted to corresponding reference process models.