Jet Infosystems will unload service center operators by more than 20% using YandexGPT

2024: Creation of YandexGPT-based services

The IT company Jet Infosystems has created a service based on the YandexGPT language model to automate the operation of its service center. This was reported to Jet Infosystems on May 20, 2024. In just one month, the company has implemented a neural network-based algorithm that can identify the topic of calls to technical support and automatically prepare responses to requests, helping dispatchers and engineers to work quickly and efficiently.

source = Jet Infosystems

The service center of Jet Infosystems provides technical support and enterprise-level IT outsourcing services. SC customers are large companies with high-load environments from all sectors of the economy: banks, retail, industry, government agencies, etc. The service center team has more than 500 engineers with expertise in various areas. Experts from Jet Infosystems are available 24 hours a day 365 days a year.

As part of the pilot project, Jet Infosystems machine learning specialists have developed an application that, using an external document repository, refers to specially created knowledge bases (for example, the knowledge base of service center operators), implements request processing scenarios and generates a relevant response. The knowledge base "Jet Infosystems" is unique, it has information about all the cases that the company's specialists have had to deal with in more than 30 years of history. Therefore, the response, which is generated using algorithms, contains the most complete information.

As part of the pilot project, it was possible to automate the processing of 30% of 13,000 requests. At the same time, 75% of these answers were completely automatic and were approved by specialists without revision. Thus, the service reduces the burden on specialists by 22%. The company will continue to improve the model and plans to implement the technology as a permanent support tool for technical support engineers to significantly speed up request processing and response quality.

YandexGPT is being implemented by thousands of companies, and there are more and more examples of large language models solving complex business problems. One example is the automation of client support, which the new YandexGPT 3 Pro model has learned to cope with especially well. The technology can accurately determine the type of call and generate a valid response based on the information provided. The peculiarity of the Jet Infosystems project is that the company managed to teach YandexGPT to work with complex IT topics in a short time, "said Alexey Dolotov, CPO YandexGPT API.

Artificial intelligence helps optimize the business processes of the 1st support line and reduces the time the client waits for a response to a request. The algorithm is already showing impressive results, and after refinement, we expect to further reduce the load of operators, - said Maxim Tkachev, director of the comprehensive support department of Jet Infosystems.

We tested several language models that are on the market and found that YandexGPT gives the best results in our project, so we opted for it. For May 2024, we are working to improve the system and improve the accuracy of its operation. According to the results of the pilot project, it shows 75% accuracy, and we want to achieve an indicator of 90-95%, - said Anton Chikin, head of the mining department of Jet Infosystems.