AI Agents (Agent AI)

Main article: Generative artificial intelligence

Agent AI is able to interact with the outside world, predict actions and act autonomously.

How agency AI works for the physical world

• The agent receives data from the environment through sensors, cameras, microphones, or APIs.

• The findings are analyzed using machine learning algorithms, neural networks or other methods based on previously trained models.

• Based on the processed data, the agent selects the actions that will lead to the fulfillment of its goal, breaking the task into subtasks.

• The agent performs the selected action using physical devices (robots, manipulators) or software (process control, content generation).

• Collect telemetry progress data for error correction.

Why agency AI is needed

• Agency AI allows you to automate complex and monotonous tasks, reducing manual labor costs. This applies to both work in the digital space and in the real world.

• By quickly analyzing large amounts of data, agent AI can find optimal solutions in real time, minimizing errors and improving performance.

• Software agents can analyze user preferences and tailor their work to selected interaction and behavior patterns.

Market estimates

2025

Denis Prilepsky: "Agenic AI is an evolution from generation to action, from automation to autonomy"

Back in 2022, we asked Denis Prilepsky pressing questions about explainable artificial intelligence (XAI), the topic that set the agenda for the digital transformation of that period. After three years of rapid development of AI technologies, we again talk with an expert with more than 15 years of experience in corporate architecture and digital transformation. This time, the focus shifts to the most relevant direction - agent artificial intelligence (Agenic AI). If XAI helped to understand how algorithms work, then Agenic AI changes the very idea of ​ ​ the role of AI in business, from a passive assistant to an active participant in workflows. Read more here.

AI agents in software development: A new era of automation of the work of programmers in Russia - TA opinions

AI agents are one of the new global trends in the development of artificial intelligence. They provide automation of routine tasks, complex information search and content preparation. TAdviser found out how AI agents are used in software development in Russia and what their shortcomings are in a conversation with market participants in August 2025.

Gartner Analysts expect that by 2028, about 15% of decisions made in everyday work will be made by AI agents, and a third of corporate applications will include such autonomous tools. According to forecasts, MTS AI the volume of the Russian market for AI assistants for developers in 2025 will amount to 17.4 billion, rubles of which 3.5 billion rubles will be spent on cloud solutions.

𝓲

Фото: Freepik

At the current stage of the development of technologies, AI agents are applicable mainly for local and limited tasks - for example, developing a separate function, writing a small module or even a component of a small service, says Andrey Malov, Product Director of TTK.Oblako at TransTeleCom. According to him, AI agents cannot yet take over the development of a complex multi-module system, and the quality of work corresponds to approximately the level of a junior developer or student: they are able to generate working code, but require constant control and refinement.

It is too early to talk about the widespread use of AI agents, agrees FabricaONE.AI Softline Nikolai Trzhaskal, director of the development of AI technologies. However, there are already examples of the use of the technology. So, Alfa-Bank he created the AlfaGen platform for the work of AI agents. They independently develop a plan for solving the problem, performs it, identifies and corrects errors, the bank said earlier in 2025.

T-Bank has developed and implemented the Safeliner AI agent to analyze vulnerabilities in code that filters false positives and generates clues that developers understand. Solutions for automatic testing and refactoring of legacy code are also being developed.

In addition, Yandex, Sberbank and MTS have functioning programming assistants: Yandex Code Assistant, which was tested by thousands of company developers, GigaCode from Sberbank with automatic code review capabilities, and Kodify 2 from MTS, which supports 90 programming languages.

According to Dmitry Demidov, head of the Norbit Innovation Laboratory, more than half of software developers use AI in the programming process, but only as an assistant who makes blocks of code, analyzes it, and answers technical questions.

The head of the BPMSoft product, Ekaterina Gerasimova, said that in most cases, AI agents in Russia still work under human control. This is due to the limitations of large language models - the risks of generating incorrect data, the nuances of contextual understanding. This is a worldwide problem, even the most advanced Western and Chinese models.

As the general director of the IT company Nobilis.Team Pavel Kalmykov said, there are known cases when agents launched into code bases with a sufficient level of privileges deleted databases, simply disabled tests if they prevented them from performing the task and many other similar cases.

Vasily Mukhin, Head of Development of Document Management Systems and Applied AI Solutions at K2Tekh, listed the main barriers to the wide development of AI agents in Russia: the lack of mature solutions, the complexity of integration and the need to restructure internal processes, as well as general market caution in matters of investment in expensive solutions.

The current situation in Russia is best described as a hybrid model, where AI agents strengthen the capabilities of developers. At the same time, we notice that gradually the market is moving from point experiments to the systematic implementation of complex and platform solutions and the restructuring of business processes for the massive use of AI in companies, "Mukhin said in a conversation with TAdviser.

The main problem lies in the overestimated expectations of users, said Alexander Karabasov, CEO of SoftMediaLab IT company. Many perceive AI agents as "virtual middle developers" who just need to give access to the repository, describe the task in a nutshell and wait for the ready-made feature. In practice, agents do not work like that, he argues.

In addition, a proactive interaction cannot be expected from the agent to clarify the details of the project - this function is still performed by a person. To obtain qualitative results, it is important to be able to form the minimum sufficient and relevant context. Otherwise, even a powerful AI can produce an unsatisfactory result, Karabasov added.

The head of the Cloud X development team, Igor Zabiyakin, sees a key problem in the presence of non-obvious errors and poor optimization in the generated code. The code looks plausible, can be successfully compiled and even run, which is certainly progress, but at the same time carries serious risks. If non-obvious errors, although quite laborious, can be tracked, then poor optimization can significantly affect the operation of the code under loads, the expert explained.

According to Vadim Soldatov, director of the Arenadata group's AI products office, it is still a little early to talk about the widespread use of not only AI agents, but also AI assistants, and the problem is not the maturity and availability of tools, but the readiness of companies to change software development processes. In addition, issues of security and confidence in the new process remain, Soldatov added.

2024

The volume of the global market for AI agents for finding customers for the year exceeded $3.5 billion

At the end of 2024, the costs in the global market of AI agents in the field of SDR (Sales Development Representative) reached $3.51 billion. About 40% of this amount fell on the North American region. Such data are provided in a study by Fortune Business Insights, the results of which TAdviser reviewed in mid-November 2025.

The SDR is a sales specialist who is engaged in the search and initial interaction with potential customers. Communication can be carried out through "cold" calls and emails, as well as through instant messengers. The goal is to determine how much a person meets the parameters of the target audience of the company, whether they are interested in the product, etc. After that, buyers who have an interest in the offered services or goods can be transferred to the sales manager who is engaged in meetings and transactions.

In the case of SDR AI agents, neural network algorithms allow you to automate the routine tasks that the corresponding employee usually performs. Specifically, AI takes over the identification and qualification of leads, personalized interactions via email or social media, and booking appointments.

One of the market drivers is the desire of companies to optimize their activities and increase competitiveness. SDR AI agents can mimic human interaction and tailor the chatroom communication process based on analysis of behavioral and context data. This level of hyperpersonalization helps to increase the involvement of potential customers, while reducing the burden on company employees when performing recurring tasks. As a result, organizations are able to scale operations without proportionately increasing headcount.

Advances in AI have a positive impact on the industry. New generation systems are able not only to support interaction through instant messengers and e-mail, but also to perform other functions, for example, to make basic calls.

The key deterrent is the need to ensure the protection of personal data. SDR AI agents handle vast amounts of sensitive information such as email addresses, phone numbers, and customer information, increasing the risk of leaks. This can lead to fines, damage to the brand and a loss of consumer confidence.

According to the AI deployment model, the market is segmented into cloud, local and other solutions (API/extensions). In 2024, the first direction dominated with 68.56%, or $2.41 billion. BFSI (banking, financial services and insurance), IT and telecommunications, retail and e-commerce, healthcare, education and other sectors (entertainment, travel and hospitality) are distinguished by scope. The BFSI segment is leading with revenue of $1 billion. Geographically, North America dominates with 40.2%, or $1.41 billion. Globally, major industry players are:

• HubSpot;
• Salesforce;
• 6Sense;
• AiSDR;
• Artisan AI;
• Luru;
• SuperAGI;
• One Floworks Technologies;
• Klenty Soft;
• Common Room;
• Qualified;
• 11x AI;
• Alta;
• Cognism;
• Gong.

In 2025, the market for SDR AI agents is expected to reach $4.27 billion. Fortune Business Insights analysts believe that in the future, the CAGR will be 23%. Thus, by 2032, costs may increase to $18.19 billion.^[1]

The volume of the global market for AI agents for the year reached $5.99 billion

In 2024, costs in the global agent artificial intelligence market amounted to $5.99 billion. The sector is booming, with more than a third of costs coming from the North American region. Industry trends are addressed in the Fortune Business Insights survey published in early November 2025.

AI agents are autonomous software systems that perform specific tasks and make decisions without human intervention. They know how to interact with the environment, collect data and based on it independently determine and perform operations that allow you to achieve pre-set goals. AI agents provide an opportunity to improve the efficiency of business operations and improve the quality of customer service. Such systems use machine learning to collect and process vast amounts of real-time data. By automating repetitive and time-consuming tasks with agent AI, organizations can accelerate tasks, improve overall operational efficiency, and reduce human error.

The key driver of the industry is the rapid development of AI in general. Organizations in a variety of fields, including healthcare, finance, manufacturing and retail, are increasingly developing specialized solutions based on agent AI to meet specific needs. AI agents are able to effectively solve tasks, observing a given model of work and adapting to changes in conditions.

On the other hand, the deployment and operation of agent AI comes with a number of difficulties. The implementation of such systems requires the collection, storage and movement of huge amounts of data, and therefore enterprises need to take adequate measures to ensure the confidentiality and protection of this information. In some cases, deep learning models may return unfair, biased, or inaccurate results, and therefore additional precautions such as manual verification may be needed. The creation of complex AI agents requires special experience and knowledge in the field of machine learning, and therefore the involvement of highly qualified workers. The training and deployment of agent AI requires significant computing resources, which is why when implementing such systems in a local environment, organizations are forced to spend large amounts of money on equipment and related infrastructure.

Technologically, the market is segmented into machine learning, natural language processing, deep learning, computer vision, etc. In 2024, the first of these areas provided the largest share of revenue - about 35.2%. Single-agent and multi-agent systems are distinguished by the type of architecture: solutions of the first type dominate. Geographically, North America leads with 34.9%, or $2.09 billion. Globally, major industry players are:

• Microsoft;
• OpenAI;
• IBM;
• Nvidia;
• Google;
• Anthropic;
• SAP SE;
• UiPath;
• Oracle;
• Zycus;
• Cognition AI;
• Mercateo AG;
• Coupa.

In 2025, the volume of the agent AI market is expected to reach $7.29 billion. At the same time, the United States will have $1.89 billion, Europe - $2.31 billion, Asia-Pacific - $1.86 billion. Latin America's contribution is estimated at about $0.2 billion. Fortune Business Insights believes that in the future, the CAGR will be 42.8%. Thus, by 2032, expenses could rise to $88.35 billion.^[2]

The volume of the global market for AI agents for the financial sector for the year reached $1.57 billion

In 2024, costs in the global AI agent market for the financial sector reached $1.57 billion. Almost half of this amount fell on the North American region. Such data are provided in a Fortune Business Insights study, the results of which TAdviser reviewed in early September 2025.

AI agents are specialized software that is able to interact with the environment, collect data and based on it independently determine and perform tasks that allow you to achieve pre-set goals. Agents can receive certain information through physical or software interfaces. Such AI is able to improve the performance of business operations by performing routine tasks. At the same time, unnecessary costs associated with inefficiency of processes, human errors and manual labor are reduced.

Banks and other financial institutions are increasingly using agent AI to automate operations, streamline processes and improve operational efficiency, the study said. AI agents allow faster detection of fraudulent activities, improve risk assessment and ensure compliance. This minimizes the number of errors associated with the human factor.

In addition, AI agents support data-driven features such as financial forecasting and credit scoring. These systems help reduce operating costs and improve the effectiveness of transaction monitoring in financial institutions. AI agents also provide customer experience improvements: they provide faster, more automated and more personalized responses. All this contributes to the rapid expansion of the industry.

On the other hand, there are also certain deterrents. Many financial institutions in developing regions continue to use outdated IT infrastructure that requires modernization and adaptation to integrate with modern technologies. However, upgrading systems comes with high hardware installation costs. In addition, the involvement of highly qualified specialists is required.

Depending on the type of solutions used, agents with conversational AI, risk management agents, fraud detection agents, credit agents, investment agents, payment and transaction agents, etc. are distinguished. The largest share in 2024 was taken by agents with conversational AI: such systems are in wide demand due to their ability to provide personalized answers and improve the efficiency of customer service. By use, the market is segmented into banks, insurance companies and non-bank financial institutions. In 2024, banks provided the largest share in the total amount of costs - 47.9%.

In terms of the deployment model, analysts consider the local, cloud and hybrid sectors. The first type of solutions dominate, which is explained by their increased level of security. From a geographical point of view, North America led in 2024 with a share of 45.9%, or $723.7 million. Globally, significant industry players are named:

• IBM;
• Accenture;
• Microsoft;
• Google Cloud;
• Cognizant;
• Verint Systems;
• KAI (Kasisto);
• UiPath;
• Darktrace;
• Sendbird;
• Cognition Labs;
• MavenAGI;
• Samaya AI;
• Affiniti;
• Greenlite AI;
• Unique AI;
• Personetics.

Fortune Business Insights believes that in the future, the CAGR in the market under consideration will be 13.7%. As a result, by 2032, expenses may increase to $4.28 billion.^[3]

Growth in companies' spending on AI agents in the world by 919% to $1 billion

Global investments in the technology of artificial intelligence agents, called "virtual employees," exceeded $1 billion in 2024, showing phenomenal growth of 919% compared to 2022. This segment has become the fastest growing technology area among 14 promising industries analyzed by the consulting company McKinsey. This was reported by analysts on July 28, 2025.

According to ppc.land, McKinsey's research is based on data on venture capital investments, M&A transactions, and public trading collected by the PitchBook analytics platform. Experts noted that AI agents quickly turned from a little-known concept into one of the most discussed areas in corporate technology.

𝓲

Фото: Freepik

In parallel with the growth of investments, there is an explosive increase in demand for specialists in this field. The number of vacancies associated with AI agents increased by 985% in 2024 alone. Such employment dynamics confirm the rapid development of the industry and the growing need of companies for expertise on this technology.

AI agents are autonomous artificial intelligence systems capable of independently planning and performing complex multi-stage tasks. These systems can interact with each other, including communication in the language they created, learn and adapt to new information without human input.

The fundamental difference between AI agents and common chat bots is their ability not only to provide the user with the result of a request, but to actively act in a digital environment and use various software tools. McKinsey analysts emphasize that this technology turns artificial intelligence from a passive tool into an active participant in corporate workflows.

McKinsey senior partner Delphine Nain Zurkia noted that AI agents are turning artificial intelligence from a passive tool into an active participant in corporate workflows. Such a transformation opens up new opportunities to automate complex business operations that previously required human intervention.

Recent technological improvements allow AI agents to perform increasingly complex tasks, including conducting research and writing program code. Systems demonstrate the ability to independently solve problems and make decisions within the framework of specified parameters.

However, McKinsey analysts state that in real business, the technology of AI agents remains largely unproven. Most implementations are at an experimental stage, and companies are just beginning to explore the practical possibilities of using these systems in production conditions.

Despite the impressive growth rate of investment in AI agents, the entire artificial intelligence sector as a whole does not occupy a leading position in terms of the amount of funds raised. In 2024, the AI industry received a total of $124 billion in investments, not including capital and operating costs of companies, which is an increase of 22% since 2022.

The green power sector remains the leader in terms of investment, attracting $223 billion in 2024. Analysts explain the dominance of this industry by the fact that energy is the basis of modern society, providing all areas - from industry and transport to digital infrastructure and everyday life.

The transformation of energy production, storage and distribution systems is seen as one of the most important tasks and opportunities of our time. At the same time, the size of investments in the energy sector shows significant fluctuations in recent years - an increase of 79% to a maximum of $315 billion in 2021, then a reduction and new growth of 14% in 2024.^[4]

Notes