TAdviser: "Artificial Intelligence: Cases and Perspectives 2025"

The conference was attended by representatives of such organizations as, Treasury of RussiaSovcombank"," Absolut Bank"," Ingosstrakh"," Credit Europe BankRussian Post"," Rosselkhozbank"," "," "," "Ural steelRosgazifikatsiya Petersburg Exchange,,," Malon Fashion Group "Beringer Ingelheim""" and Alabuga Development many others. The event was hosted by Alexey Voronin.

Business not ready to blindly implement AI

Alexey Shcheglov, Deputy Director for IT, Air Gate of the Northern Capital, listed the directions for introducing AI at Pulkovo Airport. Among such areas are video and audio analytics (analysis, event recognition, proactive reaction to them), intellectual assistants to facilitate the work of airport employees, as well as smart bots for communicating with passengers.

Alexey Shcheglov, Deputy Director for IT, "Air Gate of the Northern Capital"

The speaker dwelt in more detail on what is being done in each of the listed areas. For example, a pilot project on voice analytics at check-in desks has already been implemented at the airport. Its main goal is to monitor compliance with passenger service rules by airport employees and increase the efficiency of selling additional services.

Work on the use of AI is carried out in order to monitor aircraft maintenance, detect passenger falls on escalators, as well as to automatically adjust the climate in crowded places. The discussed uses of AI: classification of visitors by face and assignment of an internal identifier, identification of zones of interest, routes and time of stay in these zones, sex and age of the passenger.

In the direction of AI assistants, work is underway to create an assistant for the "Ground Handling Directorate" - I want to facilitate work with technical documentation - and a purchasing robot in order to robotize routine procurement operations. In addition, a system for analyzing incoming letters and the Alumni LED platform are being developed here, which would help return specialists.

We are not ready to blindly introduce AI, - summed up Alexey Shcheglov. - We are motivated by the solution of three tasks: strengthening security, increasing the comfort of passengers and obtaining an economic effect.

Dmitry Buneev, CDO, Borjomi Russia and Belarus, shared that the company has been using AI for a long time. Here it is needed to recognize products on the shelf, to work with price tags, as well as to predict production and goods. There is also your own chatbot for communicating with customers.

Dmitry Buneev, CDO, IDS "Borjomi Russia and Belarus"

In September 2024, the business began experiments to expand the use of AI technologies, and tried to do without IT specialists at all. By the spring of 2025, it became clear that the experiments did not give the expected result. But only after the IT division took control of the situation and determined the vector of development of new technologies in the company, the first notable business successes appeared.

Dmitry Buneev presented Borya, an AI assistant for employees created at Borjomi. The current functionality of the bot is the generation of on-demand reports, information support for marking processes and answers to personnel questions. He is engaged, among other things, in collecting data on line breakdowns and repair proposals based on his knowledge base on breakdowns.

The IT department has an understanding of both artificial intelligence technologies and business tasks, explains Dmitry Buneev. - As a result, we were finally able to combine all the sales data. However, after all, artificial intelligence in the FMCG segment is expensive and difficult, since we have very low margins, a lot of bureaucracy, approvals.

Banking battle over priorities

Artem Zabava, Director of the IT Development Department, FG BCS, named two main categories of players in the financial services market who actively use artificial intelligence. These are banks and big tech companies (BigTech). It is on the effectiveness of the use of AI by banks and technological giants that the financial future of the country depends, he is sure.

Artem Zabava, Director of IT Development Department, BCS

Comparing the successes in the "-iization" of both, the speaker is sure: banks successfully develop their own AI models, use them for scoring, anti-fraud, investments, chat bots of personal offers. In turn, big tech companies have huge behavioral data that AI can use to make personal recommendations. The key trend, according to Artem Zabava, is as follows: there will be a "battle" for who will become the main super-application for life and finance by 2030.

So that banks do not lose in this battle, the speaker proposed a number of specific steps. In particular, use AI not only for scoring, but also for personalizing products (making a kind of "financial Netflix"), for generative assistants ("financial GPT in the application"), and in order to improve countering fraudsters: in real time and based on deep learning technology. In addition, smart contracts with digital assets can be used.

For five years now, the struggle for the client has been going on with new means using artificial intelligence technologies, while not a good tool is enough, "said Artem Zabava. - The one who can combine data and artificial intelligence and make the correct, timely application to the client will win.

Anna Tulyakova, Head of Technology Development, AI Transformation Office, Alfa-Bank, stressed that her division is part of the bank's strategic block, so she considers the introduction of AI primarily from a business perspective.

Anna Tulyakova, Head of Technology Development, Alfa-Bank

The focus in the bank is on generative artificial intelligence, experiments with which began in 2024. In 2025, the first AI agents appeared here, and the plans for 2026 include the use of multi-agent systems. The speaker noted that both the direction of machine learning and various predictive models continue to develop at the same time in the company, and RPA remains a working tool used in many processes.

AI agents are the next step in improving operational efficiency, but we are not going to immediately abandon RPA - this is also an effective working tool - says Anna Tulyakova.

She spoke about a pilot case on the transition to an AI agent from a robot working in the structure of the operations department and controlling the targeted use of funds by customers.

The robot, which is in industrial operation, is able to qualitatively process only 20% of payment scenarios made, as a rule, by large corporate business. It is planned that a manned AI agent will significantly increase the volume of automatically processed scenarios.

Fedor Lezhnev, director of the information technology department, Alfa-Capital, said that his company uses open source AI development and external paid AI models. He outlined the AI transformation of the company in all areas: business analysis, system analysis, development, testing and support - and also highlighted the results of the transformation in each of these areas.

Fedor Lezhnev, Director of the Information Technology Department, Alfa-Capital

The speaker stressed that the focus was on employees. They were given access to external paid AI models using the Open WebUI feed. This provided a large increase in the productivity of specialists, as well as the number of ideas for the use of AI for various tasks and business areas.

We decided to use artificial intelligence in all directions at once, but the key focus is employees, - Fyodor Lezhnev shares his experience. - Already more than 60% of our employees regularly use AI tools. We regularly teach them how to use AI correctly, how to write scripts, and promote the best example.

The use of AI in sales gave savings of 50% of the budget. An increase in the number of processed client calls using AI has been achieved - from 30% to 75%. For customers, AI-generation of individual reports is carried out with an overview of the market situation, the state of portfolios, monitoring the update of financial products.

At the end of the speech, the speaker recalled that Alfa-Capital provides its clients with a mobile application, and it is the best among asset management companies. The application is used by 75% of customers, which is about 1.65 million people.

Anton Isanin, Development Director, AlfaStrakhovanie, outlined the advantages and disadvantages of generative AI, outlined an AI innovation strategy, and also gave examples of the use of generative intelligence in the company.

Anton Isanin, Development Director, AlfaStrakhovanie

Compared to classical machine learning, generative AI allows you to bring products to market much faster. Now it takes 1-2 weeks, not 6-12 months, as before. In addition, it has a wider scope, the speaker noted. However, there are also disadvantages. Among them are the high cost of machine resources and low technological sovereignty.

AlfaStrakhovanie has already put into commercial operation flexible voice analytics with quality of service control, smart document recognition (OCR), a seller's assistant for the sale of complex insurance products.

In pilot operation are mailbox reading services, as well as services for analyzing the registry of fuzzy data. The speaker named the products used in the development: a technical AI agent SourceCraft Code Assistant from "" Yandex for prompts to the encoder, MR Reviewer for analyzing code changes in the repository and for recommendations for improving the code, a generator of descriptions of technical changes and a number of others.

The focus of further innovation is the development of MCP servers (MCP, Model Context Protocol) for more efficient use of AI models, increased efficiency of AI agents, distribution in cloud approach developments, use of MCP servers in infrastructure, development of code quality control systems.

The main approach to innovation at AlfaInsurance is to subsidize the penetration of innovative technologies into processes through the leaders of changes that we identify in the process of mass training and monitoring the use of piggy banks, "explains Anton Isanin.

90% of all digital innovations - based on AI

Digitalization at EuroChem began with production as a basic function that increases the company's profit, said Nikolai Ksenzik, head of department, big data and advanced analytics, EuroChem group. He presented the results of digitalization using artificial intelligence - 2.75 billion rubles of the achieved economic effect by the end of 2025. The speaker outlined further prospects.

Nikolay Ksenzik, Head of Department, Big Data and Advanced Analytics, EuroChem

To form hypotheses on digitalization, including on the basis of large language models (LLM), the group uses problem-oriented analysis for each element of the value chain: for production, procurement, mining and enrichment, logistics, commerce, capital construction, repairs and maintenance, personnel and security. Nikolai Ksenzik gave examples of the solutions being worked out on the use of LLM, listed the planned effects.

So, for example, at the production site, they are going to introduce an operator support system for starts, stops and transitions between operator modes, as well as make support for maintaining a normal technological mode and in MR. This should lead to an increase in the speed of decision-making by the operator, an increase in the productivity of his work. Now the project is at the prototype stage, the transition to the MVP stage of the product is scheduled for November this year.

In capital construction, an MVP prototype of an AI system has been created that automates the search and assignment of engineering and procurement codes for materials and services in project and procurement documentation. This should save more than 20 million rubles a year, as well as reduce the timing of investment projects, reduce the cost of preparing and coordinating technical documentation. Further development of the product is planned starting from December this year.

Nikolai Ksenzik also spoke about a multi-agent system based on the Langflow platform, which is used by employees of business units. This system is integrated with all business applications that users work with. It can be used for processing tables, generating analytics, extracting meanings from diagrams, diagrams, graphs, converting scans into a machine-readable format, and processing documents.

The active phase of digitalization in the group of companies started in 2022, says Nikolai Ksenzik. - By the end of 2025, we are going to achieve an economic effect in the amount of 2.75 billion rubles. Almost 90% of all digital innovations are implemented in our country on the basis of artificial intelligence.

Pavel Kalmykov, CEO, Nobilis.Team, reminded the audience of the extreme congestion of the corporate IT landscape, the complexity of user interfaces and the high cost of any improvements. Headache adds a high dependence on contractors.

Pavel Kalmykov, CEO, Nobilis.Team

Any system consists of four main components: data, business logic, interfaces and integration with other systems, the speaker says. The Nobi.Chat product appeared out of the intention to simplify the construction of interfaces as much as possible: both user and integration.

During the report, Pavel Kalmykov presented the architecture of Nobi.Chat. In the center is an orchestrator agent who, at the suggestion of the user of the system, refers to the language model. It processes the request, understands what it is associated with and generates a user interface based on the provided style files and a library of standard interfaces.

The OR agent combines this interface with the data received from another component. The finished script is transmitted to the user for work. When the work is completed, the agent sends the data to be changed or deleted to the system from which it is received (CRM, ERP, etc.).

Nobi.Chat has a multimodal interface with which you can work with voice, text prompts, mouse clicks, - lists Pavel Kalmykov. - The product is focused on the execution of business logic. It independently builds interfaces, creates as an additional dialog layer around the corporate architecture, independently interacts with all the necessary components of the IT landscape.

The developed Nobi.Chat system is under piloting - in the closed circuit of the vendor and a number of customers. The speaker noted the opportunity to deploy it in the closed contour of the company for testing, and also stressed the readiness of the vendor to finalize the product model for the subject area of ​ ​ the customer.

Artificial intelligence in the days of singularity

Maxim Shpilkin, CPO for AI products for business, Sber Business Software, reflected on the basics of efficiency in the implementation of AI. First, he spoke about the company's study of AI-based products presented on the Russian market - in the industry context. There are more than 600 such products.

In a number of industries (trade, industry, healthcare), AI products are already actively used and have even become an industry standard. However, in many industry niches there are either few or no AI products. For example, in the agro-industrial complex for research and selection, marketing and marketing of products. The speaker stressed that Sberbank Business Software is ready to provide detailed results of the study to those who are interested and want details.

Maxim Shpilkin, CPO for AI products for business, Sber Business Software

Expectations of companies from the introduction of generative artificial intelligence are decreasing, the expert noted. - Technologies such as computer vision, video analytics, voice recognition, NLP, machine learning have already reached the productivity plateau. They have been tried and proven to be effective. They can work on end devices. With their help, high metrics can be achieved, and generative intelligence today is not able to completely replace everything.

Maxim Shpilkin shared the formula for the successful introduction of AI, the main term of which he called metrics. The main metric based on the results of the introduction of AI is not binary ("works-does not work"). There is much more variability here: "quality result, but long," "low quality, but fast," and so on. To achieve success in the project, you need to look for a compromise between quality and time of achievement, initially agreeing with the customer on specific metrics that are the goal.

Other components of success are qualitative representative data, pilot projects, additional training of large language models. In conclusion, the speaker spoke about several cases. So, in the airline "Azimut" an AI assistant was introduced, integrated with the booking system, for consulting customers. The assistant successfully processes most (87%) of calls in text channels. Thus, it was possible to save the costs of the first support line.

Anton Khaimovsky, head of ContentCapture, Content AI, listed the signs of singularity time in terms of artificial intelligence technology: exponential speed of change, unpredictability and "transformation of everything." AI becomes an active participant in processes, not just a tool.

Anton Khaimovsky, ContentCapture Product Manager, Content AI

We live in a time of singularity of artificial intelligence, - said the speaker. - This is a certain point in space and in time, at which it is not clear in which direction the technology will develop further, and breakthroughs in AI happen so often that you do not have time to master one technology, as the other appears.

Returning to business issues directly, Anton Khaimovsky said that generative intelligence, for example, in terms of digitization of documents, is already quite comparable in efficiency to traditional technologies (OCR). Therefore, the Content AI product implements a hybrid approach to intelligent document processing: both the classic approach (OCR) and AI are used to process unstructured content.

The speaker further shared his vision of using artificial intelligence in large corporations, emphasizing the importance of strategy, not just automating routine tasks. He made differences between AI assistants and AI agents, determined what multi-agent systems are, outlined the range of tasks that the vendor is working on for banks, the public sector and insurance companies.

So, for banks, AI agents are being developed for the automatic processing of applications for standard products (cards, deposits) in order to verify data, scoring on simple models. Agents are created to help you make a decision to automatically investigate fraudulent transactions (data collection, comparison with templates). A year ago, Content AI began developing an AI agent factory. The first release is expected in early 2026.

During the break and at the end of the conference, the participants talked informally, and also had the opportunity to familiarize themselves with the solutions and services of IT suppliers at the stands deployed in the event hall.