Russia Automates Work with Corporate Applications
In Russia, researchers have created an intelligent agent that automates routine operations inside the 1C ecosystem, helping companies accelerate workflows without compromising data security
Time Is Money
A new prototype platform developed at the MIPT Institute for Artificial Intelligence is designed to simplify how companies interact with corporate applications. The system uses a large language model and follows a “central module plus tools” architecture. The main agent generates an action plan based on a user request and delegates tasks to about twenty auxiliary modules — from document analysis to chart generation. According to the institute’s scientific director Yuri Vizilter, the solution is built for real-world use and significantly reduces processing time.
A Dedicated Tool Server
Integration with 1C is implemented through a special extension that works as a tool server. The agent strictly operates within assigned permissions: it can only access authorized data and cannot execute processes autonomously. This eliminates risks of unauthorized changes or data leaks. All processing occurs on the customer’s local infrastructure via an on‑premises large language model with about 30 billion parameters, ensuring enterprise-grade security and efficiency without transferring information outside the company boundary.
Human-in-the-Loop Reliability
Ivan Sgibnev, head of the platform solutions laboratory, emphasized that every new tool undergoes manual validation before being added to the agent. This ensures reliability and compliance with corporate security requirements. In pilot tests, the agent has already shown practical value: it analyzes tax authority requests, locates relevant documents, drafts responses, builds diagrams, and automatically extracts data from uploaded files. Additionally, it recommends actions for order processing and monitors financial operations with full contextual awareness.
