Content |
On April 3, a conference from TAdviser was held in Moscow on big data and artificial intelligence. The speakers actively discussed new business realities related to the introduction of generative artificial intelligence, the functionality of modern platforms for data management, cited cases, talked about new developments, assessed the further prospects of artificial intelligence technologies.
The conference was attended by representatives of such organizations as, State Duma,, FSBI VNII GOCHS FC EMERCOM of Russia RANEPA AstraZeneca Pharmaceuticals,,,,, Raiffeisen Bank FC Russian Post S7 AirlinesSpartak Moscow RGUNH of the Ministry of Agriculture of RussiaGoznakRitter Sport,,,,,.Severstal Pirelli
The first part of the event was hosted by Gleb Shuklin, director of the Big Data Association, the second by Mikhail Malyshev, an AI expert.
AI opens a new era in working with data
Gleb Shuklin, director of the Big Data Association, listed the potential uses of data as an intangible asset (NMA). Thus, you can increase capitalization: increase your own assets at the expense of NMA and take into account NMA in capital. You can attract investments, increase shareholder value, reduce reserves if we are talking about banks.
Data must be transferred to the "state lake" and other GIS. There is the possibility of data insurance for compensation. However, the actual state of affairs is different: big data companies have huge amounts of accumulated data, but they do not participate in the capitalization of companies, since there are no recognized methods for assessing data as assets and related data rejection practices. In addition, banks have a Bank of Russia restriction on accounting for NMA in capital, there are no recommendations and standards for accounting for data as assets.
 | Our key hypothesis: datasets are databases with all the ensuing consequences, a new object of accounting as an intangible asset, "said Gleb Shuklin. |  |
He outlined what needs to be done to make a difference. In his opinion, it is necessary to distinguish between the concept of data and related concepts (software, PAC, digital financial assets), be able to reflect data (datacets) in accounting as fixed assets (postings, value limit, revaluation), and carry out depreciation of datacets.
Boris Rabinovich, senior managing director of the data management department, SberData, is confident that generative artificial intelligence is opening a new era in data management. Progress in the development of such AI leads to radical changes in the IT industry at all levels.
The essence of the revolution is that the large-scale introduction of AI agents and assistants requires a revision of approaches to architecture and data management. Data platforms are also being rethought: tasks are being automated (copilots, AI assistants for users), a technological stack (AI-compatible data stack) is being changed. The data is processed in real time in all services.
AI transformation is based on data, so there are revolutionary changes in data management. The most important point in this transformation, according to the speaker, is the emergence of new clients of data management platforms in the form of AI assistants and AI agents. Boris Rabinovich outlined the difference between them.
An AI agent is a system based on generative artificial intelligence, capable of planning and performing autonomous actions in the external environment, responding to changes and interacting with a person and other agents to achieve their goals. In turn, an AI assistant is a system based on generative artificial intelligence aimed at automating the routine tasks of analysts and engineers, capable of performing tasks faster and with better results.
This factor creates new challenges for the data platform, the speaker emphasized. Among such calls, he called the need to implement interfaces for interaction with AI agents, repeated load growth, the need to automate typical tasks based on generative AI.
| | Now AI assistants help analysts and data engineers to perform their tasks faster, but in the near future, AI agents will take on the solution of complex tasks "turnkey," Boris Rabinovich predicted. | |
Where to get money for AI?
Dmitry Kalaev, director of Accelerator IIDF, cited data on global investments in artificial intelligence technologies. Over 100 billion were invested dollars here at the end of 2024. About 50% of venture capital investments came from AI.
However, the situation in Russia is different. In analytical reports, artificial intelligence as a separate area of investment was present for the last time in 2020. At the same time, in most Russian applied products, tools using artificial intelligence are present, although over the past two years, the growth in AI consumption by business customers in a number of segments (synthesis and speech recognition, speech analytics, chatbots, voice robots) has moved from explosive growth to moderate.
The speaker listed possible sources of investment in AI: grants, accelerators and incubators, business angels, venture capital funds (including IIDF), investments from corporations, banks, exchanges. Dmitry Kalaev gave recommendations on where it is advisable to look for investments at different stages:
- Idea stage. Only people familiar to you personally: relatives, friends, classmates, colleagues, leaders.
- First revenue. Business Angels, Angel Clubs, Venture Funds.
- Revenue is units or tens of millions per month. Venture capital funds and corporate investors.
- Revenue of more than 100 million in the last 12 months. Crowdfunding sites.
| | If the product does not use AI, then there are doubts about its compliance with the needs of the market, - said Dmitry Kalaev. - Artificial intelligence has turned into a hygienic factor. | |
Alexander Kuliev, CDO, Burger King, outlined the scale of the data stored in the restaurant chain's IT infrastructure. There are more than 100 TB of data in classic DSS (on Greenplum DBMS), 540 TB in a data lake (based on Hadoop). In addition, "Burger King" has one of the largest Arenadata DB installations in the public cloud.
| | It is necessary to move from an expert approach to data-driven work - this is the cornerstone in changing the mentality of the business. To do this, the company has developed a framework that includes uniform approaches to working with data, - Alexander Kuliev shares his experience. | |
The main approaches to analytics and working with data within the framework mentioned are that you need to constantly search for insights and test hypotheses based on big data. Do not forget about improving business models and interacting with the target audience. It becomes common practice to constantly develop, analyze, evaluate the effectiveness of the actions.
The speaker outlined the main stages of the evolution of working with data - we need high-quality data, correct data management processes, artificial intelligence in order to simulate customer behavior, and profit from business processes. He also named the levels of work with data: working with sources (correct storage and processing), building reporting (calculation and aggregation of indicators), automation of decision-making through AI services.
Alexander Kuliev admitted that the difficulties at the start of the transition to management based on the data were traditional in the company. Lack of centralized storage and data analysis, complexity of scaling, limited resources - everything is as usual. The flagship areas of further digitalization here are the "digital restaurant" (dynamic menu, personalization of offers, automation of purchases, planning the required number of employees on a shift) and the "digital guest" (attraction and first orders, registration in the application).
Irina Dolzhenko, project manager, chief expert of the informatization department, Russian Railways, said that the company is implementing more than 50 projects using AI. 28 systems use artificial intelligence. This helps to create a single information space with partners, and Russian Railways departments - to successfully fulfill their business functions.
Further, Irina Dolzhenko spoke about the AI platform, the challenges in preparing for the massive use of AI for management analytics and how these challenges were answered. Due to the lack of available data, data stores and a glossary on data analytics were developing here. The priority was on data security. To avoid romanticizing the capabilities of AI analytics, I had to resort to popularization, active search and elaboration of cases. Distrust of interpretation was overcome by the regulation of analyst responsibility.
| | To prepare the company for work with artificial intelligence and with big data, it is important to take care of the quality of this data, "said Irina Dolzhenko. - It should also be remembered that it is important not only technology, but also the training of people, the emergence of new skills in them. The most important thing here is a systematic approach. | |
Own AI agents
Before talking about the project implemented in the group of companies, Armen Amirkhanyan, Director of Artificial Intelligence Development, Moscow Exchange Group of Companies, cited forecasts of the effectiveness of using ChatGPT made by analytical agencies.
So, employees use this tool three times more often than management expects from them (according to McKinsey). BCG predicts a 20-30% cost reduction over the 2-3 year horizon, and Accenture predicts a productivity gain for ChatGPT. The speaker also cited the results of some cases. According to him, 40% of the code at Goldman Sachs is automatically written. Morgan Stanly saw a 90% reduction in lead time. The cost of processing documents at Deutsche Bank decreased by 40%.
| | All last year we introduced an AI platform in the contour of the company, this year we are reaping the benefits, "said Armen Amirkhanyan. "And we guessed with the fact that we implemented GPT, and not any little-known language model. | |
There are two options for implementing AI: either using your own GPUs based on open models, or it will be a GPU solution from vendors. The implementation process in the Moscow Exchange included the following stages. It all started with an announcement on the corporate portal, where they informed about the new technology. Then they created an active community in the corporate messenger, collected feedback, monitored user experience. It ended with a competition for the best case for using GPT.
The next step, the speaker called the introduction of the platform so that employees can make AI agents themselves. As a relevant example of such a platform, he recommended paying attention to the n8n system.
Ivan Ivanov, director of digital transformation strategy, Alfa-Bank, began by saying that business analysts used to set up processes, but now AI-based tools, the so-called "copilots" (CoPilot), are doing this. All major Western enterprise application vendors (Microsoft and others) have already created such AI tools for efficient use of the application, for employees to work in the mail, with presentations, Excel, and so on.
However, the bank does not have the opportunity to legally purchase these tools, so its piggy banks were created for Jira, Confluence, Outlook in order to use them more efficiently, which greatly changed the interaction of employees within corporate processes. The speaker spoke about how the bank moved from copilots to AI agents, which are able to completely do the work of a programmer. He noted that there are already more than fifteen such AI agents in the bank. They, among other things, are already replacing chatbots and even real operators in communicating with customers, and it is very difficult to distinguish an AI agent from a human operator.
| | About three years ago, we began to design a new organizational structure, within which people begin to interact with copilots, "said Ivan Ivanov. - We have created a Smart public piggy bank with secure access to all models. We launched the second piggy bank for developers, it works only in the inner circuit. | |
Nikita Alenkin, Head of Complex AI Solutions, MTS Web Service, spoke about the problem of describing big data inside MTS, which was solved using AI agents.
The company has more than 600 information systems, and before the project was implemented there was a lot of poorly described data, and the traditional path involving the use of analysts is very long and expensive. In this case, analysts perform routine work, analyzing dozens of sources, storefronts, understanding the structure of data. An experiment was conducted in MTS: using the Resource Inventory system for three months, 7 people processed only 310 storefronts. As a result, it was decided to create an AI agent integrated with data catalogs to automate work with data, which automatically generates a description of diagrams, tables, columns, finds critical data, builds physical and logical data models.
| | We first tried to shove the metadata into a large language model, but the result was poor. The model hallucinated, gave only basic descriptions that need to be edited a lot, says the speaker. - We followed the path of deepening data, classifying tables according to the way they are used, and then moved on to industry and domain classification of tables. And all this was built into a single business process. The system is implemented in MTS, and is already offered to other companies from the cloud in MWS. It is also possible to localize the system on the customer's servers. | |
Nikita Alenkin outlined the difficulty of creating an AI agent to obtain high-quality results of its work, since if you simply throw data into a large language model, the answer will be poor-quality. The speaker presented the technical characteristics of the resulting system: a description of one table in 60 seconds (instead of 6 hours of analyst work), and the 80% of automatically created descriptions do not require adjustment.
Not all assistants are equally useful
Together, the report was presented by Andrei Morozkin, director of operational efficiency, Finam, and Ivan Dashkevich, leading architect of AI, Finam. They talked about the stages of creating a "corporate AI brain" last year. The first stage consisted of research into the IT landscape, the needs of employees, the formation of a competence center, and easy access to large language models. It was necessary to involve employees, train and identify talents.
| | The Finam Flow platform today is like a child whom we train with data, saturate with tools, protocols, "says Andrey Morozkin. | |
The speakers listed four important components that the company focused on:
- GPT Boost - training course and hackathon;
- Chat.AI - corporate access to large language models;
- piloting AI assistants;
- Finam Flow is an ecosystem for launching AI assistants.
| | For the Finam Flow platform, we chose the "supervisor" architecture. That is, we allocate a chief assistant who delegates questions to subordinate agents, and if they do not cope, sends questions to others, "added Ivan Dashkevich. - The number of experts can be scaled indefinitely. They can be developed simultaneously, they do not interfere with each other. Right now, we're working on a self-developing system where agents themselves can change their settings. | |
Timofey Russkikh, Chief information officer, Mositalmed, spoke about some AI services: both implemented in the company and unrealized. Here they transcribed incoming calls and sent information to the service desk. Established a determination of the likelihood of returning borderline patients. Customer returns increased by 10-15%. But the doctor's assistant for filling out the medical record did not like it. "It didn't go, they considered it ineffective," the speaker explains.
The pipeline implemented in the company using AI: customer clustering, predicting customer return, individual offers, recommendations to the doctor at the appointment, in which important information is emphasized.
| | Including we used our LLM model for clustering customers in order to develop marketing offers, and this has shown its effectiveness, "Timofey Russkikh shared. | |
Evgeny Matafonov, head of basic reporting indicators, X5 Group, explained that dashboards and basic analytics have already been implemented for analysts in the company. Machine learning technologies are actively used here. The plans include the introduction of "data assistants" and decision-making by the machine.
| | Our business is actively growing, and technology must support this. In particular, there is a great need for analytics, taking into account the latest technologies, - said Evgeny Matafonov. - We studied the market and realized that many Western companies offer the necessary tools, but since they are not available for us, we decided to develop internal competencies and created our own tool. | |
The Speaker announced a new phase of data-driven business management when the system is required to respond to a routine human request. The goal of the project was to create a universal mobile assistant that gives out any information in accordance with the role model of the employee making the request.
X5 Group has developed an ICSI assistant based on the EasyReport product, which allows you to deliver analytical information both to your computer and to your mobile phone. The necessary information is provided in response to a natural language text request, according to the role model of the employee making the request.
The preliminary results look like this. The rate of receipt of information in response to the basic request is 2-3 seconds. The number of users increased by two orders of magnitude, and the number of sessions - 467 times. The tool moves to the daily assistant format.
AI will reduce the cost of intellectual labor by a thousand times
Ilya Petukhov, head of AI product development projects, Directum, stated that in the Directum RX platform, generative intelligence is already "under the hood." On the basis of the platform, more than 100 projects for the introduction of AI into commercial operation have now been implemented.
The speaker cited data from the VTsIOM study, which confirms the growing popularity of AI as a tool for solving different problems. 84% of users tried using AI. 78% of companies allocate budgets for its implementation, and 43% are already using it. Almost all respondents (97%) benefited from the introduction of AI.
Ilya Petukhov outlined the effects of the introduction of AI:
- cost reduction and/or revenue growth;
- Reduced transaction/process time
- risk leveling, identification of abnormal situations;
- reducing the proportion of human errors;
- self-learning, improving the system without direct human involvement.
In conclusion, the speaker cited the results of cases in a number of companies (SM-Clinic, Tatspirprom, UEC, etc.), confirming the effectiveness of AI technologies, and also gave recommendations on preparing for their implementation. Шаблон:Quote 'One of the stop factors for the introduction of AI is the lack of internal competencies to support the implemented AI solutions, as well as unwillingness to change anything in everyday practice, - said Ilya Petukhov.
Konstantin Egoshin, founder of the Keds Professors IT company, outlined the vital need for AI business transformation and outlined two approaches to transformation. You can either develop your own products that are compatible with AI, or you can intellectualize business processes by introducing external AI products.
Konstantin Egoshin presented a roadmap for AI transformation: adoption of an appropriate strategy, selection of AI products, architecture design, design implementation, additional LLM training, industrial engineering, integration of AI into IT, neuroscience, training and support, AI monitoring. Specialties whose tasks can be solved with the involvement of AI are a support operator, programmer, translator, personnel specialist, customer service manager, lawyer and a number of others.
| | In the coming years, artificial intelligence will reduce the cost of intellectual labor by a thousand times, - Konstantin Egoshin is sure. - At the same time, we like what is happening. Artificial intelligence will change a lot for the better. It will be easier for our economy to create added value. | |
In conclusion, the speaker cited several cases related to solving various problems: an AI assistant to a financial adviser, a tool for personalizing search, for customer support services, for video monitoring, and a number of others.
Ilya Ivanov, Development Director, Nanosemantics Laboratory, Researcher, Laboratory of Neural Network Technologies and Applied Linguistics, MIPT listed the main directions of the company's development. Here they make virtual assistants, digital avatars. They are engaged in speech technology, computer vision, analysis and generation of texts. Companies and robotics, and custom development are not alien.
Here, in total, 230 projects have been implemented. The vendor line includes ten products. The speaker spoke about some of them. So, DialogOS is a platform for creating, training and testing conversational AI. It is designed to create chat bots in instant messengers, social networks and sites, or for voice assistants in contact centers. The platform is capable of creating products that work in 40 languages, it contains 5230 dictionaries.
| | People are changing, chat bots no longer cause negativity, - said Ilya Ivanov. - It is important that the sound of the voice of the chatbot arouses the interest of the client, the desire to communicate, find out something. Then his conversion becomes more likely. | |
Ivan Mugalev, General Director, OT-OIL, listed the vendor's activities. The company is engaged in the development and implementation of IT solutions to increase the economic and technological efficiency of production processes in the fuel and energy complex (oil and gas production, well service), mining and the electric power industry. Also in the focus of her attention is the management of data and digital passports of objects, the digitalization of archives of engineering data and project documents, and, of course, import substitution. OT-OIL helps with the migration of systems to Russian software and technologies.
The speaker recalled the scale of oil production in Russia: 30% of the Russian budget is income from oil production. In 2024, 500 million tons were produced, and the total oil reserves are 19 billion tons.
The speaker listed the main goals of industry leaders: compliance with licensing obligations, reducing the cost of oil production, accidents and emergencies, and fulfilling social obligations. Approaches to achieving these goals consist of forming plans and methods of assessment, monitoring actual indicators, predicting critical deviations from the plan, changing the integrated plan.
| | Without the necessary data, no one will make the right decision - neither the state planning the country's budget, nor the head of an oil company thinking about profit and fulfillment of social obligations, nor an ordinary employee who needs to perform certain actions - lists Ivan Mugalev. - And here the basis of success is industry expertise in working with data, creating the right models based on it with a view to the future. | |
In conclusion, the speaker spoke about the ATOLL platform - functionality, business effects of use, technology stack (open source, PostgreSQL DBMS).
During the break and at the end of the conference, the participants communicated in an informal setting.
| The site content is translated by machine translation software powered by PROMT. The machine-translated articles are not always perfect and may contain errors in vocabulary, syntax or grammar. Read original article If you find inaccuracies or errors in the results of machine translation, please write to editor@tadviser.ru. We will make every effort to correct them as soon as possible. |