This paper proposes an architectural and technological recommendation for constructing a call bot to access the possibility of interaction with generative AI in the form of large language models via standard GSM-transferred phone calls. In an effort to establish a loose template to guide business-oriented call bot implementations, a real-time voice-processing pipeline was established and thoroughly described using a selection of commercially available AI-based technologies and their interconnection, resulting in the creation of a prototype application. The paper further explains the roles of used technologies and their usage, illustrates the used architecture, and evaluates our solution’s actual business-case success performance compared to openly available generic call automatons.