Využití umělé inteligence v podnikové praxi – vytvoření chatbota pro zákaznickou podporu

This bachelor’s thesis describes the feasibility and effectiveness of using generative artificial intelligence (LLMs and RAG architectures) to automate customer support in an e-commerce environment. The main outcome of the thesis is a fully functional AI chatbot, designed and deployed in the live environment of the Czech e-shop jawavysocina.cz. The system was developed on the low-code Voiceflow platform with integration of ChatGPT 4o-mini and the Pinecone vector database, thereby achieving an 80 % reduction in product recommendation latency compared to competitors. Based on live testing from February to March, the chatbot successfully and independently resolved 83 % of routine inquiries at the L1 support level without the need to escalate to human support. Furthermore, an analysis of real conversations confirmed high user satisfaction, with 81 % of customers who rated the chatbot giving it 4 or 5 out of 5 possible points. An economic evaluation of the implementation demonstrated direct savings in human resources and immediate profitability with a return on investment (ROI) of 81 %. This real-world implementation thus demonstrates that AI tools are now both technically and financially accessible to small and medium-sized enterprises, providing them with a measurable competitive advantage.