Russian companies have postponed or closed about 90% of pilot projects for the implementation of generative artificial intelligence, launched in 2025. This is reported by the newspaper Vedomosti, citing a survey by the consulting company "Intellectual Analytics." The study was conducted from December 2025 to February 2026 among 50 of the largest organizations in the fields of IT, industry, finance, the public sector, transport and logistics. Only 7–10% of projects related to large language models, AI-based chatbots, and AI agents reached the stage of full-fledged industrial implementation. The remaining 90% either remain in piloting, are transformed, or are completely closed.
Why AI Implementation Projects Didn't Take Off
According to experts, about 30–40% of pilot projects were curtailed because they did not give the expected financial effect. The main reason is weak integration into the real business processes of companies. In most cases, models are not integrated with CRM (customer relationship management system), ERP (a comprehensive program for managing all business processes), document management, and other corporate systems. Another factor is that companies launched not practical solutions, but high-profile initiatives with a strong PR effect.
Examples from the survey:
- One company independently retrained the Chinese model Qwen for the legal department without involving external experts. The data was collected in insufficient volume, the accuracy of the AI assistant did not exceed 30%, the project was closed.
- Another organization wanted to implement AI in the support service, but most requests contained documents and images. The models available at that time did not have the necessary multimodality — the project was curtailed.
AI Spending and Schedule Shifts
Typical budgets for pilot projects for the implementation of AI (excluding IT infrastructure costs) in 2025 ranged from 5 to 15 million rubles. Half of the surveyed managers postponed the transition of pilot projects to industrial operation from 2025 — early 2026 to the second half or the end of 2026. The main reasons: higher than expected implementation complexity, the need for staff training, IT infrastructure modernization, and information security issues.
At the same time, according to experts, the level of AI tool adoption and the culture of working with them in companies is 80–90%, while the real integration into business processes is only 5–10%. Barriers: risks of leaks and abuse of models, lack of corporate culture and a shortage of specialized specialists.
The mass curtailment of 90% of projects does not mean a loss of interest in AI — rather, the market is moving from "hype" pilot projects to conscious implementation. Generative AI is capable of giving an additional GDP increase of up to 2.5% by 2035 (about 46.5 trillion rubles), but this requires not separate "wow" solutions, but systemic integration, restructuring of business processes and training of personnel. The shift in deadlines to 2026 and later suggests that "normal hard work" is just beginning.
Read more materials on the topic:
- The Ministry of Digital Development has published a draft law on sovereign AI: neural networks should be created in Russia and trained on Russian data
- Rosatom will significantly increase the use of AI as part of its digital strategy until 2027
- The introduction of AI into work will help Russian companies save up to 4 trillion rubles a year
