Rostelecom Contact Center Doubles Speech Recognition Accuracy With AI
Machine learning and large language models are reshaping how customer support operates, helping Rostelecom deliver faster, more accurate responses while easing the workload on human agents.
For several years now, Russian telecom operators have been using artificial intelligence to streamline workflows, reduce employee workload, and save time on routine tasks. These technologies are proving particularly effective in corporate call centers.
AI Agents Work Within a Unified System
By the end of 2025, the efficiency of Rostelecom Contact Center had increased significantly. The company deployed a full suite of AI-driven solutions. A key focus was the integration of modern machine learning (ML) models, large language models (LLMs), and AI agents. Together, these technologies automated a wide range of routine processes and saved hundreds of hours of staff time.
The operator integrated AI agents into a single system where they operate in parallel and complement one another. As a result, the share of tasks handled by AI agents has doubled.
Rostelecom Contact Center also adopted robotic process automation to handle standard and predictable tasks. Activities such as transferring data between systems, processing documents, verifying information, filling out forms, and generating reports are now performed automatically.
More Than 20 Million Requests Per Year
ML and LLM technologies, combined with advances in speech synthesis and recognition, have significantly improved the performance of voice and text-based bots. Over the course of the year, speech recognition accuracy doubled. At the same time, AI assistants communicate so naturally that 15% of customers believed they were interacting with a human agent. In total, more than 23 million requests were processed using speech analytics in 2025.
“We replaced the contact center’s interactive voice response system with a new intelligent assistant. As a result, the share of issues resolved on the first contact exceeded 80% for the first time, and this figure continues to grow. We will keep deploying proprietary innovations and tools to achieve even stronger results,” said Alexander Svyatets, Director of Rostelecom Contact Center.
Automated Service Quality Assessment
The company also introduced a new system for evaluating customer service quality. Dialogue analysis and topic classification are now fully automated. When needed, the system generates recommendations for operators in real time. This provides deeper visibility into business processes while improving service quality and reducing costs.
The modernization has also changed the structure of customer interactions. Most users now resolve their issues through text-based channels. The number of processed calls in the mass segment declined from 26.9 million in 2024 to 24.5 million in 2025 (-9%), while chat requests increased by 7%, from 7.8 million to 8.5 million.
A Result of Systematic Transformation
These results reflect several years of consistent modernization efforts at Rostelecom Contact Center. Back in 2023, its communication analytics system was registered in the national software registry. That same year, it processed more than 1.2 million calls and over 1.7 million chat requests.
In 2024, the contact center began integrating a large language model into its operations. At the time, the expectation was that LLMs would enable automated call evaluation using checklists, improve understanding of natural language from both customers and agents, and provide more human-like guidance.
The results now confirm that this long-term strategy is delivering measurable gains. Customers benefit from higher-quality support, while the operator has automated many processes and significantly reduced costs. AI adoption at Rostelecom is expected to expand further and eventually cover all core operations.
