Applications of Generative Artificial Intelligence: Chatbots in e-commerce

The bachelor thesis deals with the use of generative artificial intelligence for the development of advanced e-commerce chatbots and their integration into existing online communication channels of e-shops. The thesis first provides a brief overview of the evolution of chatbots from historical systems such as ELIZA to current large language models such as GPT. It then compares and analyses different options for implementing artificial intelligence in chatbots, focusing on the comparative creation of a custom large language model, fine-tuning of already pre-trained models, as well as the use of a RAG system that appropriately combines the generative capabilities of artificial intelligence and the ability to retrieve relevant information from a knowledge base. Based on this comparison, a RAG system that effectively combines generative capabilities and enterprise knowledge base retrieval is selected and discussed in detail for specific e-commerce needs. Following this choice, requirements and functionalities for the chatbot are defined. The design of the chosen RAG architecture and its practical implementation into a working prototype of a business assistant for an e-shop is described in detail. Finally, the thesis is dedicated to the evaluation of the solution used and its comparison with a human operator in terms of key metrics: cost, response speed and quality of the provided answers.