- Статьи
- Economy
- Using someone else's mind: industrial companies invested in generative neural networks for the first time
Using someone else's mind: industrial companies invested in generative neural networks for the first time
Large Russian industrial companies have begun investing in generative neural networks, market participants told Izvestia. The tools previously used to write texts and create videos without human intervention have become necessary in factories for managing 3D printers, developing new products, and more. For example, R-Pharm creates medicines with their help, the Russian Copper Company develops project documentation, and Norilsk Nickel trains employees. How much the introduction of such networks costs enterprises and how it affects production efficiency is described in the Izvestia article.
Why do industrial enterprises need neural networks
Russian industrial enterprises are already actively experimenting with generative neural networks, says Sergey Dutov, Director of Corporate Innovation at the Skolkovo Foundation (VEB.RF Group). They should not be confused with artificial intelligence, which businesses have been implementing for 10 years, he emphasizes. Currently, such neural networks are used in complex decision support systems.
— For example, this is a metallurgist's assistant, when, based on an assessment of the quality of incoming raw materials, technologists are asked to change the setting of metal smelting equipment. It's not just a search, it's collecting information from diverse sources, generating ideas and recommendations on how to apply them," he explained.
Another example is the use of neural networks in procurement procedures.: she can not only find the necessary part among the internal databases and warehouse balances, but also recommend its equivalent.
Another area of application of generative neural networks is 3D printing. In this still promising scenario, the generative neural network will be able to "invent" itself, for example, a new engine or parts for it, and immediately print a sample, added Sergey Dutov.
Currently, at least 20 Russian industrial companies use generative neural networks in one way or another.
For example, in addition to the metallurgist's assistant, Norilsk Nickel uses their opportunities to train employees, according to materials from the Skolkovo Foundation, which Izvestia has reviewed. R-Pharm is developing technology that allows using such networks to quickly develop medicines (to design models of molecules of substances), and the Russian Copper Company is developing project documentation. Izvestia sent inquiries to these organizations.
NLMK is actively developing the field of generative artificial intelligence, the company told Izvestia.
— Generative neural networks are used to automate workflow, train employees, support production staff, and create digital assistants. A special focus is on the development of AI solutions: intelligent assistants for searching the knowledge base, automation of the support service and the introduction of AI into internal development processes. In 2025, we plan to achieve economic benefits for several pilots and solutions in circulation by optimizing business processes," the company's press service told Izvestia.
Severstal uses generative neural networks to increase production efficiency (increase unit productivity, reduce costs, monitor equipment operation), improve the quality of finished products, monitor compliance with safety regulations, and reduce environmental stress, said Andrey Golov, head of the Severstal Digital Center for Artificial Intelligence and Machine Learning.
— In addition, it is being implemented in sales and corporate processes. The total project portfolio currently includes over 60 solutions. The use of technology is already yielding significant results — in 2024, the economic effect of projects in this area amounted to over 1 billion rubles, " he told Izvestia.
The editorial board sent requests to the Ministry of Finance.
How AI improves production efficiency
The Ministry of Industry and Trade records an increase in demand for such tools, especially for the use of such tools locally, within a limited sample of industrial enterprise data, the agency told Izvestia.
There can be many scenarios: for example, generative AI analyzes data from equipment sensors (temperature, pressure, vibrations, and other parameters) and predicts potential malfunctions. This reduces the risks of unplanned downtime and extends the service life of the equipment. Generative AI models are used in design and modeling, and generate designs based on input parameters (material properties, cost constraints, and production methods), the Ministry of Industry and Trade listed. Generative neural networks are also needed to accelerate the development of software products, they added.
This technology is increasingly being used in industry and in business in general, said Leonid Konik, partner at ComNews Research. It allows you to automate repetitive operations, from generating financial reports to preparing standard documents for issuing a loan. The neural network is able to quickly "view" and compare thousands of documents — a person either cannot do it at all, or it will take him a lot of time, he explained.
— Regulatory regulation in almost any area of the economy is multifaceted and often contradictory. For example, Roscosmos has recently been using neural networks to identify inconsistencies in GOST standards, and there are many such documents in the rocket and space industry," the expert said.
Neural networks are ways to bring (and are already bringing) a huge economic effect, repeatedly speeding up the result and eliminating errors, Leonid Konik is sure.
— In our experience, the introduction of a generative artificial intelligence-based assistant in industry can be divided into three stages. The first is a quick hypothesis test based on customer data.: It costs several million rubles," MWS AI (part of MTS Web Services, formerly MTS AI) told Izvestia. — The second stage is a full—fledged pilot: about 15-20 million rubles, the duration is one and a half to two months. The third option is scaling to the entire function or company: 35-45 million rubles. There may be several such projects in the company.
According to the data provided by Pavel Komarovsky, director of business development of this company, at the CIPR-2025 forum, neural networks save seven hours of time every month for each employee of the administrative unit only to search for the necessary documents. On an annual basis, based on the cost of man-hours, savings amount to 1.3 million rubles per department of 10 people. You can save 19 hours (3.8 million rubles) on analyzing these documents, and so on. For a company with about 2,000 office staff, annual savings could reach 3 billion rubles, he estimated.
— Applied generative neural networks can multiply either the net profit or the revenue of an enterprise as a whole, or they can both, depending on how they are applied. Neural networks of the most popular type of LLM (large language models) can be used to improve internal processes, for example, through the creation of a digital double of an employee or manager," said Anton Averyanov, CEO of the ST IT group of companies, TechNet NTI market expert.
The introduction of generative neural networks is positive not only for the industry, but also for the end customers of its products, says Denis Kuskov, CEO of Telecom Daily. Huge savings in time and money lead to lower prices, he explains. In addition, new products are appearing on the market, such as medicines or non—original 3D-printed machinery parts, he explained. This is very important in the light of the policy of import substitution and sanctions, the expert concluded.
Переведено сервисом «Яндекс Переводчик»
