multi-agent systems – a step towards strong artificial intelligence / Economic news of Krasnoyarsk and the Krasnoyarsk Territory / Newslab.Ru

There is an active paradigm shift in the financial industry: from simple RPA bots to intelligent agents that interact with each other through special protocols, and with humans using natural language. AI agents are valuable for their ability to significantly speed up routine operations and increase the efficiency of business processes, but algorithms are still a long way from truly human-like intelligence (AGI). This was stated by Sergey Golitsyn, head of T1 AI (part of the T1 IT holding), speaking at the session “Independent AI agents in the financial market – utopia or reality?” within the framework of the Finopolis-2025 forum.

“Until recently, simple neural network models were launched manually, but now we have transferred interaction with AI agents to a native language format. Thanks to this business, it was possible to simplify the work with disparate data, automate its processing and speed up operational processes several hundred times. However, this is still just an innovation in automation. The AI ​​agent does not have a fundamental understanding of the world and its own goal – it is an algorithm that operates within given parameters,” explained Sergei Golitsyn.

The expert called the next step in the evolution of artificial intelligence the transition to multi-agent systems in which AI agents will be able to interact with each other, clarify tasks for each other, and rebuild analytical models.

“As a developer, we are moving on to building a chain of agents who not only act as executors and transmit the results of their own analysis to other agents, but interpret the data in their own way and, if necessary, change the parameters of the models. This is a meta-level of coordination that allows for improved decision making,” he said.

Golitsyn emphasized that multi-agent systems will have a strong impact on the corporate sector. They allow you to combine data from various sources – storage, BI systems and cloud services – and conduct complex analytics.

At the same time, the expert noted, the development of such architectures requires special regulation. Since multi-agent solutions will directly affect the financial stability of organizations and trust in digital systems, it is necessary to develop standards for their testing, monitoring and reliability assessment.

“We are moving towards strong artificial intelligence not through an overnight breakthrough, but through architectural development. Multi-agent systems are the key to a new level of artificial intelligence,” summed up Sergei Golitsyn.