Citizen letters (requests, complaints, recommendations) addressed to the Government were being circulated and processed inefficiently in the Mulberry document management system, slowing decisions and reducing the productivity of public institutions.

We partnered with leading AI researchers at DeepPavlov (Moscow Institute of Physics and Technology) and the system’s developer VxSoft to embed real-time, automated document classification directly into the workflow public officials already use. Pilots ran across three high-volume institutions.

• An AI algorithm for real-time automated classification and labeling of
citizen correspondence
• A centralized collaboration model between Mulberry users and the system developer
• Ministry-level pilots in Health, Education, and the Prime Minister’s Office

Increased operational performance, measurable productivity gains, and a working blueprint for embedding AI into existing public administration software, without forcing institutions to change platforms.