Сбер's Strategic Expansion into Agentic AI

The strategic direction within Sber involves transitioning certain LLM capabilities toward practical utility—first through high-value niche applications like investment analysis via GigaChat, and now towards autonomous office routine management using new benchmarking standards.

Key Development Milestones

• Financial Intelligence Integration (GigaChat): A new tool in SberInvestments allows users to analyze asset price fluctuations against real exchange events; this targets both professional time-saving and novice onboarding/education.
• New Benchmarking Standards (Cowork Bench): The release of Cowork Bench establishes a way to evaluate how well AI agents handle actual work processes across mail, spreadsheets, and corporate systems based on 475 end-to-end tasks.
• Context vs. Model Capability: Preliminary tests indicate that successful automation success depends more significantly on the software managing context rather than just the underlying model itself.

Note: Reports confirm platforms such as GigaCowork have already demonstrated strong performance with a 75.2% task completion rate or higher under these conditions.

Bottom line: Sber is building an integrated ecosystem where specialized financial intelligence meets generalist agent ability, prioritizing context-driven automation for real-world workflows.

! DYOR (Do Your Own Research)