"AI engineer" will help Beeline save specialists' time and improve the quality of communication.

      On March 26, 2026, at 12:29, Beeline is implementing artificial intelligence technologies in incident management processes on the network. Pilot tests of the "AI Engineer" project are currently underway in several regions of the country. According to forecasts, the implementation will allow for a twofold acceleration in the time taken to address incidents and a 5% reduction in average problem resolution times.

      The operation of communication infrastructure is impossible without incidents — that is, unplanned events that disrupt the functionality of base stations or the transport network. This can include, for example, cable breaks, equipment unavailability due to power supply issues, equipment hangs, or issues with individual boards. Information about the failure is transmitted to the operator's engineers, who determine the cause and resolve the problem either remotely or on-site. Now, Beeline engineers will be assisted by artificial intelligence in incident management.

      The development called "AI Engineer" combines ML models trained on Beeline's knowledge base, RAG knowledge (i.e., context from verified external sources), customer instructions, equipment signal information, and incident history. Upon receiving a message about an incident, the "AI Engineer" summarizes the context, translates natural language into command chains, performs diagnostics, and either initiates active actions on the equipment (for example, where a reboot is required) or, if the problem cannot be resolved by its means, sends structured information to a human engineer. At the initial stage, the "AI Engineer" remains under the control of a specialist; however, for certain types of incidents where accuracy exceeds 80%, we have already launched active agent actions and minimized re-checking. As accuracy increases, the volume of manual verification decreases.

      Thanks to its broad context and continuous retraining, the "AI Engineer" successfully handles even unknown failures. In the future, developers plan to expand its functionality through predictive analytics capabilities — to anticipate and prevent network failures.

      The project's goals are to relieve specialists of routine tasks that cannot be automated by classical methods, to capture a digital footprint (the AI can log all its actions without slowing down the process, unlike a human who records only 5-10% of critical moments due to time constraints. People often miss symptoms, reasoning logic, and even causes of incidents, resulting in a loss of valuable material for retraining the AI), to reduce the number of on-site visits, and to lower downtime metrics. In other words, to free up resources, increase accuracy and responsiveness — and ultimately improve communication quality.

      Valery Shorzhin, Deputy General Director of Beeline for Technology, stated: "About 40% of incidents occurring in Beeline's network are typical, which means they can be automated. Standard automation has been in use for a long time, but its capabilities are limited: it is difficult to rigidly program an action scenario when interaction with external systems is required or when flexible responses are needed based on context. Therefore, we created a special AI agent.

      In just over two months of pilot testing, the new system has processed about 15,000 requests. The availability rate was a stable 99.9%, and no significant technical failures were recorded. The average decision-making time for an incident was only three minutes. This architecture can easily handle any volume of incoming tasks while maintaining stable processing times. Cloud infrastructures have also proven to be reliable. We are not standing still; we regularly run new open-source LLMs through our internal telecom benchmark. For example, we are now using a fresh line of Qwen models in the AI agent. Our goal now is to expand the application of this solution across the entire Beeline network."

      The NIA "Nizhny Novgorod" has a Telegram channel. Subscribe to stay updated on major events, exclusive materials, and operational information. Copyright © 1999—2025 NIA "Nizhny Novgorod". When reprinting, a hyperlink to NIA "Nizhny Novgorod" is mandatory. This resource may contain materials 18+.