Inteligentní procesní automatizace – využití jazykových modelů ve firemních procesech

Robotic Process Automation (RPA) is an established tool for automating repetitive business activities, but it has a number of limitations. Intelligent Process Automation (IPA) extends RPA with elements of artificial intelligence, with current developments directed primarily toward the use of large language and multimodal models. The aim of this thesis is to map approaches to integrating large language and multimodal models into runtime RPA/IPA solutions and to verify the feasibility of a selected approach through an original prototype. The thesis employs a systematic literature review in the form of a scoping review, focused on current scholarly studies from 2025–2026. It includes a thematic synthesis of selected studies according to the role of the language model in automation, the method of its technical integration, and the limitations of traditional RPA addressed by this integration. The practical part follows with the development of a prototype that verifies the use of a language model as a tool within a predefined workflow for classifying customer requests and extracting structured data from synthetic telecommunications inputs. The contribution of the thesis lies in the creation of an overview framework for understanding current possibilities for using language and multimodal models in intelligent process automation. The thesis may serve as a basis for further research as well as for the practical design of solutions focused on working with unstructured inputs, the adaptability of automated processes, exception handling, and the long-term sustainability of RPA/IPA systems.