The fourth edition of the IT Elements conference opened in Moscow on September 9 under the theme “Business Continuity 2.0,” emphasizing organizations’ ability to maintain operations after cyberattacks, infrastructure failures, or major outages rather than simply relying on recovery plans.
Special attention on the first day was given to ransomware. According to Jet CSIRT, ransomware accounted for 69% of confirmed incidents in the first half of 2026, with the majority of attacks occurring in the second quarter. Participants examined how to safeguard backup copies against compromise, why declared RTO values often diverge from actual recovery performance, and who should make critical decisions when infrastructure is already down.
Another major topic was artificial intelligence. Experts discussed corporate AI agents and reminded attendees that even the most advanced models remain unreliable without high-quality data. Companies were advised to strengthen Data Governance, Data Quality, and DataOps practices before deploying AI at scale.
AI workloads are also transforming data center infrastructure. A joint study by Jet Infosystems and AC IKS found that more than 30% of surveyed organizations have already created dedicated network segments for AI, while another 15% plan to do so. Customers are requesting racks rated at 80 kW and 120 kW, with some inquiries reaching 200 kW. Robotic systems managing power consumption can reduce data center OPEX by up to 30%.
On the networking track, participants reviewed import substitution efforts. Testing of Eltex and EcoRouter devices demonstrated sufficient performance in typical scenarios, yet no single domestic solution was found to cover all use cases. Equipment must still be validated against specific traffic patterns and architectures.
A separate discussion addressed workforce implications. AI is taking over routine tasks from junior specialists, making career progression less linear. The market increasingly needs professionals who understand business requirements, work across disciplines, and can critically challenge model outputs rather than simply reporting them.


