AI agents are taking over the economy: bankers and marketplaces revealed at the St. Petersburg International Economic Forum when robots will replace us.

      The moderator was the first deputy chairman of the board of Sberbank. Representatives from the largest banks, a well-known marketplace, the chemical industry, and an IT giant participated in the discussion. The global market for AI agents has almost reached $10 billion and is expected to grow 15 times by 2034, according to forecasts. Experts explained that these tools are similar to familiar Excel but can not only analyze but also act—communicate with other agents, negotiate, and adapt to feedback. Gradually, they are becoming independent market participants. A survey among session participants showed that one in three believes that by 2030, AI agents will make 40-60% of decisions in their companies without human involvement. The spread of technologies is hindered by two things—the speed of progress and regulation. In the future, there will be both universal agents and specialized ones for specific tasks. Bankers noted that AI agents sharply reduce information asymmetry and eliminate intermediaries. For example, an agent can independently browse hundreds of websites, select the best offer, negotiate, and provide the user with a ready-made solution. A person will not have to search for anything. At the forum, a platform was presented where agents from different companies can make deals in real-time through blockchain. However, the main question for businesses is how to get into the circle of trusted agents of the client. According to bankers, trust comes to the forefront. There will be thousands of agents, and a person will not be able to use them all. They will choose those they trust in various areas. Companies will thus turn into "commodities"—standard products. It is important what metric the agent will maximize: immediate profit or long-term customer happiness. The competition for loyalty will reach a new level, reports transsibinfo.com. A representative of the marketplace added that so far no one has created an agent that independently navigates different platforms, compares, negotiates, and purchases on behalf of the user—this is a kind of "Holy Grail." Two barriers hinder this. The first is trust: people are just learning to use built-in assistants and are not always ready to rely on external ones. The second is that people love to choose for themselves; they enjoy comparing and making decisions. Therefore, the process will not be quick, but the company wants to become a pioneer and create a first-choice assistant. Bankers also predicted that the agent economy will make financial conditions between banks practically identical: everyone will have access, quickly and seamlessly. But people are not as rational as they seem (as proven by Nobel laureates). Therefore, banks are transitioning from "fintech" to "fanthech"—who provides more pleasure, with whom it is pleasant to associate oneself. For the next five years, this will become the main factor of competition. A representative of the chemical industry spoke about three areas of implementing AI agents in companies. The first is transactions for the back office. The second is the production system, which has been imported and is now integrating agents for the complete reproduction of technological processes. The third, completely new, is when a company's agent interacts with a counterparty's agent. All three tracks are actively developing. Changes in the labor market were discussed separately. A representative of the IT company reminded that every 10-20 years, a technological revolution occurs. Initially, programming was done on lamps, then languages, compilers, low-code, and no-code appeared. Now the programming language is the human language. The number of programmers has never decreased; it has only grown—the entry threshold has lowered, and the market has expanded. With the arrival of agents, one programmer will be able to launch a factory of agents and do the work of ten. In the future, everyone will be able to manage such a factory and do many times more. In conclusion, the moderator suggested that the audience assess the economic impact of the agent economy by 2030. Almost half of the participants responded that it would exceed 20%.