Comindware, a Russian developer of low-code platforms, conducted an internal study on AI adoption in business. The research indicates that while companies are comfortable delegating routine tasks to AI, they maintain human oversight for financial decisions, deal qualification, and revenue forecasting.
The company has updated its CRM AI agents to follow a human-in-the-loop model. Agents analyze data from sources like DaData and the Federal Tax Service to propose solutions, but managers must confirm these actions before they are finalized. AI is permitted to act autonomously only for non-sales related tasks, such as routing technical support queries.
CEO Vitaly Shpak noted that because AI cannot be held legally or financially liable for errors, human oversight remains essential for regulatory compliance. This approach aligns with current Russian legislation, including Federal Law No. 243-FZ, which took effect on September 1, 2026, and clarifies that legally significant decisions must be made by humans.
Comindware expects future industry regulations to formalize the human-in-the-loop requirement. The company is currently designing its products to ensure transparency and human control, anticipating that trust in AI will grow over the next one to two years as companies accumulate performance statistics.
