"Basis", the largest Russian developer of solutions for managing dynamic IT infrastructure, has implemented neural network technologies in business processes related to incident response and product technical support. The use of generative artificial intelligence has reduced the time spent on routine operations by 20-50%, and the speed of completing related tasks has increased up to 6 times. As a result, labor efficiency has increased, and the company's customers have received higher quality and more prompt service.

The solutions developed by Basis specialists are based on open language models that can understand natural language, analyze large volumes of technical documentation, and generate program code. They are integrated into the company's existing workflows and function as digital assistants, taking on routine operations. The operation of the neural network - including its training and application (inference) - is organized in an isolated, secure environment. All tools are located within the company's own infrastructure - this approach ensures complete control over the data and eliminates the risks associated with using external services. 

Key implementation results

Reducing incident response time. The time that the responsible engineer spent organizing work in the event of an incident was reduced by 70%. Previously, he had to manually form working groups, create communication channels, and prepare initial information about the problem, but after the introduction of neural network technologies, these processes were automated. As a result, all the necessary specialists receive structured information about the incident as soon as possible and can immediately begin to eliminate it, which is critical to ensuring the uninterrupted operation of services.

There are plans for further development of this area: the digital assistant will be responsible for accounting and controlling the incident resolution process, maintaining its chronology, it will "participate" in the discussion and analysis of logs, preparing a draft report and a retrospective on the incident.

Instant information retrieval. An intelligent assistant (digital tech lead) has been created on the basis of the neural network, which instantly finds answers in the corporate knowledge base and provides ready-made solutions. The system analyzes user requests, finds relevant information among hundreds of technical documents, and generates accurate answers with specific instructions. It is especially important that the assistant successfully processes requests even with vague formulations, understanding the essence of the problem and suggesting optimal solutions. It also provides the user with a thought process and a list of links to the documentation used to generate the answer. This allows the specialist to notice nuances that the assistant may have missed, or to come to a deeper understanding of the problem.

Accelerating the development of scripts and tests. Routine tasks for writing simple code, which previously took hours of work by qualified specialists, are now completed in 10-15 minutes. Productivity in solving these tasks has increased by 3-4 times. The neural network has taken over the writing of scripts, the creation of tests, and the analysis of existing solutions. 

"The introduction of neural network technologies has become a logical step in the development of our company. We are not just following technological trends - we are creating practical tools that really improve work efficiency. And over the past six months, we have achieved significant results, including thanks to colleagues from the "Public Cloud" RTK-TsOD, who helped us with the organization of infrastructure. Gradually, we are moving from simple automation of routine tasks and preparation of documentation to the integration of neural network technologies with our other systems and processes. This will improve the efficiency of our work and, as a result, provide our customers with more prompt and high-quality service," saidDenis Romanov, Director of the Professional services Department of Basis.