Agentic AI: business process automation in SMEs through artificial intelligence tools

This thesis examines the role of Artificial Intelligence (AI) in the automation of business processes in small and medium-sized enterprises (SMEs), with an emphasis on the use of Agentic AI and Retrieval-Augmented Generation (RAG) architectures. There is created a comprehensive Agentic AI system (technological artifact), inside an automation platform (n8n), employing a Design Science Research Methodology. The aim of the thesis is to transform the theoretical capabilities of modern ΑΙ systems into practical and functional digital tools capable of supporting core administrative functions of SMEs such as Sales, Marketing, Finance, and Administration, reducing the time spent on repetitive, multidimensional and cognitively demanding tasks. This method allows the thesis to address specific research questions such as designing and developing Agentic and Agentic RAG architectures, measuring the performance of AI agents through KPIs, and picturing AI agents as digital "role assistants" aiding different organizational functions. This thesis follows the Design Science Research (DSR) methodology, proposed by Peffers et al. (2007), with the aim of forming a structured methodological framework for the development of innovative Information Systems. This process allows for the systematic identification of the business problem, the definition of the design objectives, and the translation of the theoretical objectives into an applicable development framework, which is then used for the design, implementation, and evaluation of the proposed Agentic AI system. More specifically, the methodology is applied on a real operational context of a SME in the Used Car industry, which is used as a starting point for the analysis of the requirements for an AI system implementation. As a result of this process, a structured framework is developed that can be followed by a wide range of SMEs for the deployment of innovative Information Systems. Furthermore, following the Digital Transformation framework described by Doukidis et al. (2020), the research illustrates in what ways Agentic AI systems have the potential to facilitate Business Process Transformation and Organizational Transformation if intelligent agents are integrated into main managerial workflows. In addition, there are designed multiple specialized conversational AI agents based on RAG architectures, in order to connect with the company information, which operate autonomously to support Managerial decision-making across different Organizational Departments, such as Sales, Marketing, Finance, and Administration. Moreover, a central Management AI Agent is created, which serves as the only point of interaction for all the Managers, directing questions to the relevant functional agent and providing consistent and well-documented answers. Finally, there is carried out a quantitative performance evaluation of the system that demonstrates how Agentic RAG systems enhance not only operational efficiency but also decision-making quality.