2026-05-30T08:39:04.075Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:1:11427 2025-06-13T23:06:51Z 1

The Economics of AI-Powered Call Center Development Using ChatGPT for the Needs of an Automotive Retail Business 2025 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/1/11427 https://portal.sinteza.singidunum.ac.rs/Media/files/2025/Sinteza-2025.pdf P. Dakić T. Heričko Ž. Kljajić V. Todorović Amid the shift toward automation in customer service through artificial intelligence (AI), this research study explores the practical integration of OpenAI Realtime API using the \texttt{GPT-4o-mini-realtime-preview-2024-12-17} model into an AI-powered call center tailored for the local automotive webshop. Based on OpenAI's documentation, this study investigates the cost structure, technical implementation, and integration techniques of a ChatGPT language model through ChatGPT Realtime API, a streaming API suitable for real-time interactions with token-by-token response and low latency, within a PHP-based environment. The study aimed to develop a scalable, multi-lingual AI-powered call center prototype leveraging WebRTC and PHP to deliver both online, i.e., text-based, and phone, i.e., speech-based, customer assistance. The system employs WebRTC for real-time speech and text communication, while PHP facilitates seamless interaction with existing databases and backend systems. The prototype focuses on system integration, real-time data processing, and efficient API connectivity to improve automation in the context of online auto parts retail. The resulting prototype represents a cost-effective alternative to traditional call centers, offering faster response times, reduced operational expenses in terms of staffing expenses, and improved customer experience. conferenceObject 382 388 https://doi.org/10.15308/Sinteza-2025-382-388 https://portal.sinteza.singidunum.ac.rs/Media/files/2025/Sinteza-2025.pdf