NLMK IT conducted an experiment migrating a legacy MES system to a modern stack using a team of AI agents coordinated by a single engineer. The pilot project involved a system managing shipping documentation for six production workshops with over 20 external integrations.
The use of AI agents reduced IT development labor costs by 10 times during the initial phase. The total cost of the full migration was approximately 5 times lower than the original estimate for a traditional approach.
Despite the cost savings, the project timeline remained at six months. External system integrations and user-related tasks, such as training and feedback analysis, could not be fully automated and required traditional methods.
The migration process allowed the company to formalize system knowledge previously held by individual experts. This data now powers an AI assistant integrated into the new system, which currently handles user queries and support tasks.
The company aims to resolve 70–80% of user requests through AI agents without human intervention. Future plans include enabling these agents to analyze errors and propose technical fixes for human approval.
