Interactive Voice Response (IVR) systems are a cornerstone of modern call centers. But anyone who’s navigated a phone menu knows the frustrations: endless loops, confusing options, and the desperate hope of reaching a human. Traditional IVR struggles to handle the increasing complexity and volume of customer interactions. This research explores how AI can transform IVR, focusing on the unique challenges of the Kazakh language. The key lies in leveraging the power of large language models (LLMs). The researchers integrated automatic speech recognition (ASR), text classification using LLMs, and speech synthesis to create a more intelligent and efficient IVR. One of the hurdles was adapting this technology to Kazakh, a language with nuances not well-represented in standard datasets. The team tackled this by fine-tuning the OpenAI Whisper model with a specialized Kazakh speech dataset, significantly boosting accuracy. They used Low-Rank Adaptation (LoRA), a technique that improves model performance without dramatically increasing its size. For the text classification itself, they employed the IrbisGPT model, chosen for its understanding of Kazakh. A key innovation is integrating Retrieval-Augmented Generation (RAG), further enhancing the model’s knowledge base. The practical implementation of this system in a call center saw impressive results. The AI-powered IVR dramatically reduced the burden on human operators, freeing them to handle more complex issues. Customers benefited from faster, more accurate responses, and the call center saw overall efficiency improvements. This research highlights the transformative potential of AI in call centers. Future development could focus on refining the system's understanding of complex requests, integrating more advanced dialogue management, and adapting the technology to other under-resourced languages. The move towards AI-powered IVR paves the way for more personalized, efficient, and satisfying customer experiences.