Voice and chatbots in Russia and the world: volumes, dynamics, drivers, barriers

Summing up and forecasts: volumes, dynamics in the Russian Federation and the world

According to the consulting agency Grand View Research, in 2023 the global chatbot market reached $5 billion, and in the next few years, growth is predicted to continue - up to 23% annually. By 2028, the global market could reach $15.5 billion. The Russian market is growing faster: according to Just AI analysts, in 2023 its volume was at 33 billion rubles. According to general estimates, annual growth is 30%, which is higher than the global pace. And in five years this figure may exceed 80 billion rubles. Vladislav Viryasov, director of Avantelecom, relies on such estimates in his analysis. He also pointed out that in the field of client service, about 76% of contact center managers plan to increase investments in AI tools in the next 2 years. At the same time, 37% of companies already use such technologies in the field of text and voice communication channels with customers.

Director of Artificial Intelligence Sber Business SoftMaxim Ivanov cited a third-party estimate that the volume of the chat bots market in 2023 amounted to $6.7 billion, and by the end of 2024 it is expected to grow to 8.43 billion. By 2028, the volume could reach $22.42 billion, with a CAGR of 27.7%.^[1]

Roman Milovanov, head of the development of chatbots and voice robots at Satel, said that in 2023 the global market for voice robots and chatbots continued to grow, and it is expected that by 2027 its volume will reach about $100 billion with an annual increase of 19%. And in Russia in 2023, the volume of the voice assistant market also increased by 30% compared to 2022, reaching 4 billion rubles, as shown by the Smart Ranking study. The conversational AI market in Russia will grow at a rate of 40-80% per year, and by 2025 its volume will amount to $561 million, according to estimates by Just AI.

The specialist noted that in the next 3-5 years the market will continue to grow dynamically, there will be a full introduction of voice assistants in the field of education, health care and the business sector.

The global chatbot market is showing steady growth, said Nikita Murenkyi, head of the Kode dialog product development team. The main drivers of global market growth, according to the specialist, are the introduction of AI into business processes, the development of machine learning and the growth of automation of most processes, and the most active global industries are e-commerce, healthcare, fintech and the public sector.

Many companies will actively invest in technologies such as natural language processing (NLP), generative AI and self-learning models to improve the quality of communication. The specialist is confident in the prospects of the trend for hyperpesronalization in user experience. The share of voice in managing digital services will grow along with the growing up of the audience, which since childhood has become accustomed to interacting with voice speakers and perceiving audio information. This trend, according to the expert, will become especially bright in 5-6 years.

There are already successful examples of the use of AI assistants in smart city systems and educational platforms. In Russia, taking into account the emphasis on the digitalization of public services, chat bots will be actively introduced into the systems of municipal and federal services.

The field of voice and chatbots for business with the development of AI turned out to be inextricably linked with this technology industry.

The study McKinsey shows that generative AI can bring the global economy up to $4.4 trillion per year, and according to Yakov & Partners, the full economic potential of AI in Russia by 2028 will amount to 22-36 trillion rubles in nominal prices, while the realized effect by 2028 can reach 4.2-6.9 trillion rubles, which is equivalent to an impact of GDP up to 4%. Generative AI accounts for about 20% of the realized effect - 0.8-1.3 trillion rubles. Such data were provided by the Vladislav Belyaev executive director and co-founder of the AI platform. AutoFAQ

According to the expert, the potential of generative AI lies in the plane of automation of work tasks, which occupy 60-70% of the working time of employees. And if such tasks as writing texts, creating images using neural networks lie on the surface, then few people pay attention to the fact that employees spend about 20% of their working time a month looking for information among the accumulated information in the company. This problem can also be solved with the help of GeneAI.

Positive market drivers

As Svetlana Zakharova, Business Development Director of Just AI, said, Russian developers are quickly finding cases of using new technologies, actively investing in their products. This, in turn, stimulates venture capital funds and investors, who, against the background of all political events in the country, did not understand where the market was moving and where to invest. Against this background, more and more startups and AI enthusiasts appear who want to move the industry forward.

According to Vladislav Belyaev, an unequivocally positive factor for market growth is that thanks to LLM, the speed of creating a number of chat bots and intellectual assistants has grown very much, and it has become possible to launch projects that were previously economically impractical.

Vladislav Viryasov noted the significant steps that Russian developers have taken over the year in the development of dialogue logic platforms. Many market participants have released updates or new advanced systems in terms of integrating natural language and machine learning, speech recognition systems.

Alexander Pavlov, Managing Director of Robovoice of SL Soft, mentioned the growing interest of companies in digitalizing processes due to a shortage of personnel, as well as an increase in consumer requests for the level of service. Also, according to the expert, the development of technologies plays an important role. Modern bots are capable of performing more complex tasks, including consulting, accepting orders and even selling.

At the same time, their performance is indeed comparable to the level of employees, "he stressed. - This opens up new opportunities for business and contributes to the further growth of the industry.

An equally significant positive factor, according to the expert, is state support and initiatives: programs to stimulate digitalization contribute to the introduction of modern solutions and market expansion.

Tatyana Gaponenko, Marketing Director at Nanosemantics Group of Companies, also noted the positive impact of state interest. According to her, in addition to the general interest in AI technologies, there is a direct indication of implementation from the state, and, as a result, there is a great demand from customers who want to use them in business.

"Voice" technologies made a "transition" along the diffusion curve of innovations from "innovators" closer to "most": the development of technology led to a gradual decrease in the entry threshold for automation of client service in the natural language, and the emergence of LLM attracted business attention to speech technologies in general, although the models themselves have not yet reached the level of confident application in the Enterprise segment. Such an examination was shared by Alexander Krushinsky, Director of the Voice Digital Technologies Department of BSS.

Russians are largely techno-optimists and well accept new technologies, believes Nikita Murenkyi. According to him, in the Russian Federation there are very successful products on the mass market, their own solutions for the development of conversational AI, competing with the world ones. All this motivates business to test bold hypotheses.

According to Alexander Sidorov, leading engineer of the AMT Group contact center department, the company's experts believe that the main driver for market participants is the desire to follow modern trends, as well as statistics of successful implementations, which clearly shows what economic effect can be obtained from implementation. An important factor is the improvement of the attitude of customers to robotic systems, which are increasingly common in everyday life.

Maxim Ivanov noted that one of the important breakthroughs was the Retrieval Augmented Generation (RAG) technology, which allows you to create chatbots without classic fine-tuning, trained on the company's knowledge and using LLM to generate responses.

Anna Panina, an expert in the practice of chat bots K2 Tekh believes that the market drivers should include the technological development of smart assistants, who are now able to understand requests in a simple "human" language. And users appreciated it. From customer support to collecting summary information from different sources for decision-making, such as tendering, there is a growing desire to use them everywhere.

I don't want to call them just bots anymore, "the expert said.

According to her, the growth driver must recognize the ability to put all the required components and modules in the internal circuit of the customer. And, in addition, the availability of smart assistants for more customers in different functional areas expands flexible mechanisms to distinguish between access rights to processed information and information that is transmitted in response to requests.

Proposal structure

The market for voice and chat bots of the Russian Federation can be called oligopoly. This is evidenced by the comments of experts. However, estimates of supply saturation vary.

Vladislav Belyaev said that there is a situation on the market when on the one hand there are two large companies that are developing their LLM - Sber and Yandex, and on the other hand there are companies that provide ready-made products based on Open Source models and have been working in the market for support of both external and internal processes for several years. And then there are the developers of rules-based chatbots, which in the current realities significantly lose to AI solutions, which makes them a small business tool that cannot yet afford to implement intelligent solutions.

As AutoFAQ, we belong to the second group, - said the expert. - At the same time, Yandex and Sberbank solutions were initially aimed at V2S, but we see cases and potential competition with them in the B2B market.

He stressed that there are companies on the market that prefer to engage in their own development with internal resources. But when choosing such a path, business must take into account all risks.

So far, many are opting for a "self-written" digital assistant on AI technologies due to the low entry threshold for creating a prototype. But the creation of an industrial fault-tolerant solution that will pass all information security checks is a much longer story. While all the processes are lined up, big business will lose time, and the market and technologies will go ahead - the medium-sized business will not have enough money for its own development. Therefore, if we correlate temporary and financial risks, then the ready-made solution has clear advantages here, summed up Vladislav Belyaev.

Svetlana Zakharova also noted that there are two oligopolies on the market - Sberbank and Yandex, which compete for the ability to produce and develop their LLM solutions. According to the expert, companies from industries that previously either did not depend on technologies or almost did not focus on them are starting to come to the market. For the most part, this includes large industrial companies that are starting to use genia to automate back-office processes. Also on the first line are companies that have the ability to create products based on genia.

Just AI is one of the leaders in this direction, "Svetlana Zakharova emphasized.

Although the number of players in the market is large, there are still few products with support for advanced technologies based on LLM, says Alexander Pavlov. According to the expert, the market structure includes several large players in leading positions, but there is also space for the growth and introduction of more innovative bots, especially with a focus on the use of LLM to improve the quality of user interaction.

The expert noted the fragmentation of the market supply, which has become a noticeable trend. Players develop niche bots for industries or tasks. This allows the business customer to quickly obtain specialized solutions at the optimal price.

Nail Akhmedzhanov, technical director of ELMA Bot believes that the market cannot be called a rich offer, since domestic solutions do not reach foreign ones, and it is not entirely appropriate to use the latter now.

If we talk about large language models (LLM), then in the Russian Federation there are two market leaders - these are Yandex and Sber, - the expert noted.

Шаблон:Quote 'An oligopoly has developed on the market with the dominance of large players such as Sberbank, Yandex and VK, - said Roman Milovanov.

According to the specialist, each of these giants stands out for its unique strategies: Sberbank focuses on manufacturability and ecosystem, Yandex offers a wide range of products integrated with the Alice assistant, and VK focuses on user loyalty and integration with the social network. Roman Milovanov noted that the presence of other participants in the market, such as Tinkoff, speaks of high competition and saturation of the industry.

In general, the market is quite saturated and continues to saturate, says Alexander Krushinsky: for each of the elements of the IT landscape there are several vendors, some of which are focused on the Enterprise segment, and some on SMB. Large enterprises typically look for a complete solution that can be installed in the customer's closed loop with all architectural stack and integration requirements met. Small contact centers, without specific information security requirements, and in the absence of their own IT staff, often choose cloud solutions in which the most knowledge-intensive tasks (speech recognition and synthesis, LLM) are closed by cloud solutions, for example Yandex SpeechKit.

There are offers on the market for some and others, - said the expert.

Vladislav Viryasov proposed a slightly different approach to segmentation. The expert said that currently there are about two dozen players on the market offering solutions in the spectrum from typical analog bots to complex customized voice models.

In general, Russian vendors can be divided into those who specialize exclusively in the introduction of voice or text robots, and those who offer automation solutions as a separate product in an entire ecosystem of communication solutions, he said.

Demand structure

As Vladislav Belyaev said, for 4-5 years AI bots have been actively introduced in industries such as telecom, retail, the banking sector, online education and online services.

Now we see a trend in the integration of AI-based chatbots in industry, construction, power and government services, "he said.

According to Vladislav Viryasov, the financial sector and the banking sector are one of the main consumers of "dialogue" technologies. Here chatbots are used to serve customers, automate consultations and process requests. The expert also noted a strong increase in the medical industry, where virtual assistants are being introduced to record citizens for an appointment with a doctor and provide information on the services of medical institutions. Among other growing areas, he noted real estate, where AI helps improve the automation of communication with customers, organize the search for objects and answer frequently asked questions. Also, chatbots are being actively implemented to support users, configure product recommendations and process orders in the retail and e-commerce areas. Stable growth, according to the expert, continues in the public sector in terms of automating the interaction of citizens with municipal and departmental services through dialogue interfaces.

Maxim Ivanov noted sales, tourism, medicine and production. The construction industry, according to the expert, is just beginning to master bots, while development is already deepening in production.

Alexander Pavlov singled out as the main medical and financial spheres, retail and logistics. According to the expert, companies in these markets are introducing "dialogue" technologies both to automate client service and to optimize HR processes for hiring staff, organizing, and information support for employees.

Nail Akhmedzhanov said that the most susceptible to "dialogue" technologies were industries with the highest degree of digitalization, namely retail, banks and telecom.

In general, the spread of voice and chat bots is dictated by the size of the contact center and, as a result, the economic effect of automation, said Alexander Krushinsky. According to him, the largest CCs, in addition to telecom, banks, retail, are also in the federal executive bodies (FOIV), and logistics, medicine and insurance companies are slightly less relevant.

We have also recently observed the penetration of "dialogue" technologies into internal service processes - IT/HR support for employees and recruiting processes, - said the expert. - But in share terms, they, of course, occupy a small share of projects to automate customer service.

Tatyana Gaponenko believes that the financial sector still retains leadership in the use of AI technologies, also maintains high rates of retail and e-commerce, and significantly more implementations in the insurance industry and medicine.

For example, in 2023, in one of the projects, Nanosemantics specialists improved IVR using a voice chatbot, which allowed the client to significantly reduce the load on the contact center ‒ 40% of customers are now served by a voice robot.

In Russia, dialogue technologies are actively used in the banking sector, telecommunications, retail and healthcare, said Roman Milovanov. According to the specialist, over the past year there has been a significant increase in their use in public services and education to automate processes and provide information to users, and industry is deepening its use in the field of production process management and maintenance, which allows improving interaction between operators and systems, as well as increasing the effectiveness of monitoring and diagnostics equipment.

Nikita Murenkyi also believes that the highest financial results in the Russian growing market of spectrum technologies are shown in the banking sector, retail and the public sector.

SMB and Large Business Demand Differences

The segment of medium and large business usually prefers craft solutions configured for specific tasks and with additional adjustment of systems, said Vladislav Viryasov. Preference may be given to products with advanced features, for example, integrated with existing CRM and other systems, offering chatbot personalization for customers, data analysis and knowledge management functions, natural language processor (NLP) and machine learning.

Large and medium-sized businesses focus more on the maturity of the architecture, the security of the solution, as well as its scalability, Maxim Ivanov shared. Medium and large companies facing more complex tasks choose omnichannel platforms. Such advanced solutions support various communication channels, offer expanded functionality for creating and managing chat bots, noted Roman Milovanov.

The larger and more complex the business, the deeper the request, and the tougher the product requirements. If we talk about Enterprise, then the AI system to support customers or employees should have high fault tolerance, fully comply with the company's information security requirements, be able to install On-Premise, include the hybrid AI + human model, be able to integrate APIs with internal systems, for example , Service Desk, CRM , BI , MES, etc., comply with the project KPI, as well as be in the register of domestic software, noted Vladislav Belyaev.

Large customers are focused on creating an ecosystem in their contour, so they are ready to allocate specialists and hardware resources. This deployment model allows you to get integration with all adjacent systems and expand the range of tasks that the robot solves, said Alexander Sidorov.

Large companies, which often have inside the AI division, use our NLab Marker, a platform for creating datacets using AI. This is a professional tool for quick markup of images, audio, texts and medical images, - shared Tatyana Gaponenko.

According to the specialist, large and medium-sized companies choose the DialogOS dialog platform for creating and training voice and text robots, and natural speech recognition (NLab ASR) and speech synthesis (NLab TTS) technologies are in demand by all companies, regardless of business size.

Experts said that small businesses usually prefer simple solutions, such as voice robots, which automate basic tasks without complex configuration. Small companies, as a rule, choose boxed and inexpensive solutions and are ready to work with startups on Open Source. Most often, they are interested in basic functionality with automatic answers to frequently asked questions and the ability to integrate with several communication platforms at once. They often choose cloud solutions with a subscription licensing model, because in this case, no computing costs are required, there is no need to train personnel to administer the new system, and you can stop using the product after the end of the license period.

Interest and benefit of the state

Serious state support for digital technologies has a very strong impact on the technological development of the Russian market, said Anna Vlasova, head of the computer linguistics department at Nanosemantika Group Nanosemantics.

As part of the National AI Development Strategy, the state finances startups and allocates grants. An index of the readiness of industries to introduce AI into their business processes is also being formed. Flagship companies in their industries are also interested in maintaining leadership in this area, which stimulates them to use AI-based solutions.

Experts note the unconditional interest of the state in the development of the direction. They talked about the implementation of successful projects in health care, education, social services and other areas.

In medicine, voice and chatbots are used to automate the process of making an appointment with a doctor, providing medical information and monitoring patient health. In education, they help in the creation of interactive educational materials. In the MPSC, on the regional websites of public services, bots help in managing citizens' requests and providing services.

It is important for the state that business finds new growth points for itself, and in this regard, the development of geniia is the same point, "Svetlana Zakharova believes. - It is also important to solve the problem of personnel hunger, and automation using geneAI can help solve this problem.

We work with a number of state-owned companies and services in customer support processes directly and by providing the Xplain AI Copilot tool to help contact center operators, "Vladislav Belyaev shared.

Vladislav Viryasov considers the introduction of a national strategy for the development of AI in Russia to be one of the main drivers of the market. According to the expert, the interest of the public sector is growing every year, and today the state considers the development of AI technologies as one of the main development priorities. They are planned to be used to improve public services, services and institutions, such as unified regional contact centers, medical organizations, MPSC, employment centers. The efficiency of providing the service, the availability of government services, the lack of expectation when contacting government agencies are priority areas that are implemented using voice and text assistants, the expert said.

Alexander Pavlov noted that government agencies are interested in introducing such solutions to improve efficiency in various areas, including education and social services. In medicine, chatbots and voice robots are also used, and for a long time. Their functionality is already beyond standard tasks, such as accepting appointments or making an appointment with a doctor.

The interest of state organizations is caused by the need to automate routine processes, says Alexander Sidorov. The specialist spoke about the frequent misconception of customers who believe that if necessary, turning any public service into a digital service, it is enough only to change access to it.

This is fundamentally wrong, since in order to ensure high-quality digitalization, first of all, it is necessary to transform the service itself, that is, the business process, - said the specialist.

Alexander Pavlov also said that recently there has been an increasing need for unambiguous identification of AI when interacting with citizens. In particular, this is due to the wide coverage of the population by various social networks and services, when accessing which users must unequivocally understand who they are communicating with (who is "on the other side" of the screen).

It should be noted that in the Russian Federation there is a so-called "Code of Ethics in the Field of Artificial Intelligence," indicating the need to identify AI when communicating with a person, but the effect of this code is currently exclusively advisory, he said.

The next reasonable step, in his professional opinion, here is the participation of the state in the transition from ethical rules to legal norms in human rights issues in critical areas of life (for example, the mandatory provision of the possibility of termination of communication or other interaction with elements of AI at the request of the user).

Market unavailability factors for breakthrough development

The main deterrent to the spread of new technologies is the high cost of On-Anticipate projects, because full-fledged work requires appropriate specialists and expensive equipment. This opinion in the answers of experts sounded most often. In addition, professionals pointed out the need to take into account legislative restrictions, the negative impact of overly high business expectations, the need for a large amount of well-structured data for training chatbots and a number of other nuances.

The public sector and large business want to place the solution in their contour, most often this leads to a significant increase in the cost of projects, - said Maxim Ivanov.

Nail Akhmedzhanov noted that high-quality language models require large computing resources, in particular expensive GPUs. And recently, customer companies have tended to deploy all solutions at their own capacity, but running LLM locally is not a cheap pleasure.

When customers see the price, interest in advanced artificial intelligence quickly cools down, the expert shared his experience.

Alexander Sidorov commented on this issue as follows: for customers, the main difficulties are the amount of hardware resources that need to be allocated for the implementation of AI products, as well as the lack of specialists within the company who could independently support these systems. According to him, AI is still complex, obscure to many technologies, but its benefits are obvious, and after a while it will be applied in many products and in all industries.

Vladislav Belyaev believes that the combination of information security requirements for placing On-Anticipate solutions and the high cost of server capacity with video cards is a significant factor restraining the development of the direction. This increases project costs and implementation times. To solve this problem, AutoFAQ offers customers a number of workarounds, including working in the cloud only for data processing (without storage) or running generative AI functions in a limited mode on conventional processors.

Among other limiting factors, the expert noted the disappointment of a number of companies at the start of the trend, since at that time the bots being created were still low. Other companies in the wake of the "hype" around ChatGPT placed great expectations on LLM and, having received unsatisfactory results, lost faith in the effectiveness of the tool.

In these conditions, we always return the customer to the main issue - to solve what problems he is looking for a product, - Vladislav Belyaev shared the solute.

As practice shows, companies do not always need to implement LLM, but a product based on traditional AI based on question classifiers, not answer generators, is enough.

By traditional AI, we mean the ability of the system to understand a person in order to give the right answer, run a script, translate to an operator or give him a hint (prompter), the expert made a reservation.

Roman Milovanov also pointed out the difficulties associated with high user expectations. In addition, in his opinion, the situation is influenced by high competition, the need for innovation and regulatory barriers, difficulties with the integration of services.

To overcome these barriers, it is necessary to increase the volume of investments in the creation of new products, carefully monitor legislative changes and market needs, study foreign experience, the specialist recommended.

The business has formed overestimated expectations from intellectual solutions, especially from solutions based on generative AI, confirmed Anna Vlasova. According to her expert opinion, overestimated expectations arose as a result of powerful information pressure, because the media (and not necessarily specialized industry ones) write about generative AI, they talk about it at forums and conferences. Also, the state not only financially, but also administratively stimulates business to digitalization. And all together this creates a picture of an "ideal assistant" in any area, and even self-learning. This is even true: AI can be applied in many fields, and it learns from algorithms that are embedded in the model.

But at the same time, AI ‒ this technical solution. This is mathematics, not telepathy, - Anna Vlasova is surprised at the stereotypes. - This is a solution with a lot of advantages and huge potential, but not without a number of restrictions.

One such limitation, she said, is AI's reliance on the quantity and quality of the data he trained on. If the training data is provided by the customer company, then the trained AI will work as a spotlight that will instantly highlight all the shortcomings in this data: unstructured or poorly structured documents, storage mixed with current and outdated versions of documents, contradictions in internal knowledge bases, and so on.

Another limitation (but it is also a feature, and the advantage of AI ‒ depends on which side to look at) ‒ is the introduction of intellectual "fuzziness" in decision-making. In other words, the logic of generative models, different from traditional search, must be taken into account when introducing AI into companies' business processes.

Take the trend of this year: searching for documents or knowledge bases using a chatbot based on generative AI, - said the specialist. - Even with the most stringent settings, the generative model may not be free from excessive "creativity" or "hallucinations" (in the terminology of AI specialists).

As Vladislav Viryasov said, chat bots work better in natural language, so until now there may be difficulties in understanding various accents, jargon or slang or the context of the conversation, especially when it comes to complex topics. In addition, chatbot training requires large amounts of data, and the need to comply with the emerging new regulatory requirements imposes additional restrictions. In addition, until now, the reality of many companies is technological limitations and a lack of competencies and knowledge, it is difficult for them to attract qualified specialists and maintain the relevance of their technological stack.

Just AI expert Svetlana Zakharova said that the expectations from 2024 among specialists in the number of cases implemented and the speed of implementation were much higher, but, as practice has shown, the market was not ready for such a rapid development of gene AI technologies. The main brakes on progress were the following factors: ensuring safety when working with LLM, opaque pricing of LLM solutions, unavailability of existing solutions or processes for the implementation of LLM, lack of competent personnel, low immersion in technologies (at the moment), difficulties with calculating KPI when implementing solutions based on GIA, poorly developed issue of implementation and application from a legal and legislative point of view.

Companies are just beginning to find options for how to order hardware, many are just now getting money for their own data centers or pilots of LLM solutions, the expert frankly shared. - There is definitely no dead end development, since there is still work on the preparation of high-quality data sets, training and further training of models for specific business needs. We are all just at the beginning of this journey.

Notes