An AI assistant for officials has appeared in Russia. Civil servants discuss issues at a meeting, and the system then compiles a meeting protocol from this. The Ministry of Digital Development has already included the modular platform in the register of Russian software.
The platform's peculiarity is that it is considered Russian software, but internally it relies on open technologies Ollama and ChromaDB, created by foreign developers. The first is responsible for text generation and processing, and the second is used as the system's long-term "smart memory." At the same time, the platform is deployed completely locally, within the closed circuit of the department.
The solution developer is the Russian IT company OOO "Incase Corp". This is a small firm with an annual revenue of just over 100 million rubles.
The company is engaged in the implementation and support of electronic document management systems based on Docsvision and SDU "Prioritet". Its clients included the Ministry of Sport, the FMBA of Russia, and other government agencies. In addition, the company has a separate direction related to its own developments based on artificial intelligence.
The system can upload audio or video of a meeting, and also connect to a video conference. After that, the AI recognizes the voices of the participants, separates their remarks, and matches them with the real names of employees from the departmental directory.
Then the neural network cleans up the transcript: it removes repetitions, interjections, and inappropriate expressions. Colloquial speech is translated into a strict official business style. After that, the assistant summarizes the meeting and identifies specific instructions from the conversation: what needs to be done, who is responsible, by what deadline, and which department is involved. The output is a ready-made draft protocol in DOCX format.
At the same time, the documentation emphasizes: AI does not make decisions instead of officials and does not evaluate whether they are correct or not. It only works with what was actually said and recorded at the meeting.
The assistant also has "memory" of past meetings: thanks to the vector database, it can search for necessary solutions and materials not only by exact phrase, but by meaning, in plain language.