Nornickel Saved Tens of Billions of Rubles: How AI Agents Optimize Budget and Project Control

Nornickel, one of the world leaders in AI implementation in industry, has saved tens of billions of rubles by integrating artificial intelligence into its processes. By 2030 alone, the company plans to increase this figure to 50 billion rubles. AI agents, developed in collaboration with Reksoft, have enabled the automation of budget and project control, reducing design times from months to days and increasing decision accuracy. In enrichment, neural networks already control 70% of processing units and can make decisions 100 times more frequently than human operators.

In a large industrial company, budget and project management involve a colossal amount of data, manual reconciliations, and approvals, which consume hundreds of thousands of working hours. Every stage, be it planning, control, or reporting, is subject to human error, leading to delays, cost overruns, and reduced overall efficiency. These are invisible costs that accumulate and become a significant expense item, but this problem is solvable today.

The Scale of the Problem: Where Billions Were Lost

Nornickel is a giant structure with continuous production cycles, complex logistics, and a vast number of projects. Before the implementation of AI agents, budget and project control required significant human resources. Planning, data collection from various sites, budget approvals, tracking project execution – all involved manual information processing, numerous communications, and lengthy approval cycles.

This approach led to several critical problems: long decision-making times, a high risk of errors due to human factors, inability to react promptly to changes, and consequently, significant financial losses. The issue was particularly acute in design, where timelines could stretch for months, and inaccuracies led to cost overruns during construction.

Why Traditional Methods Fell Short

Previously, the company used standard project and budget management systems, but they could not provide the necessary level of automation and adaptability. These systems were suitable for structured tasks but struggled with the enormous volume of unstructured data, complex interdependencies, and the need to make real-time decisions. Humans had to manually reconcile data from different sources, analyze it, and make decisions, which was too time-consuming.

Nornickel concluded that a fundamentally new system was needed—one that could not just store data but actively interact with it, analyze it, and propose optimal solutions. This led to the idea of AI agents capable of performing the functions of a comprehensive intelligent assistant.

How AI Agents for Budget and Project Control Were Designed

The task was ambitious: to create a team of AI agents capable of taking on the functions of a general designer and ensuring end-to-end budget control at all stages of project lifecycles. The agents were designed as a complex system consisting of several modules, each responsible for its own set of tasks.

The first agent focused on planning and budget optimization. It collected data from all departments, analyzed historical indicators, forecast costs, and proposed optimal resource allocation scenarios. The second agent specialized in project control: tracking task execution, monitoring deadlines, identifying deviations, and warning of potential risks. The third agent handled design automation: from preparing technical specifications to issuing final documentation, using its proprietary domain model, MetalGPT.

A key feature was the agents' ability to work with unstructured data, understand context, and make decisions based on comprehensive analysis. They were integrated into the existing IT infrastructure to minimize the need for employees to change their workflows.

Implementation and Scaling

Implementation proceeded in stages, starting with the most critical and resource-intensive areas. Initially, AI agents were launched in pilot mode to optimize design processes. This allowed for quick evaluation of effectiveness and necessary adjustments. Subsequently, the agents' functionality was extended to budget control, and then to monitoring production processes and even developing new materials.

The Nornickel team actively trained to work with the new tools, and AI agents gradually took over routine operations, freeing up specialists for more complex and creative tasks. An important aspect was that AI did not replace people but became a powerful assistant, increasing overall productivity and accuracy.

Results of AI Agent Implementation

In addition to direct savings, Nornickel achieved significant qualitative improvements: increased planning accuracy, reduced errors, accelerated decision-making processes, and the ability to react promptly to changes. AI agents have become an indispensable tool in conditions of qualified personnel shortages, compensating for the lack of expertise at remote production sites.

How to Implement This in Your Company

Nornickel's case demonstrates that AI agents can be a powerful tool for companies of any scale, especially in industries with large data volumes and complex processes. To begin implementation, consider the following steps:

• Identify routine and resource-intensive processes. Find areas where employees spend a lot of time on repetitive operations, data collection, and reconciliation. This could include budget control, project management, logistics, or customer support.
• Start with a pilot project. Choose one relatively small but critically important area for AI agent implementation. This will allow you to test the technology, evaluate its effectiveness, and get initial results.
• Develop a domain model. For AI agents to work effectively, a specialized knowledge base adapted to your industry's specifics and business processes is crucial.
• Integrate AI agents into existing systems. The less employees have to change their habits and learn new interfaces, the faster and more successfully the implementation will proceed.
• Train your team and scale. Gradually expand the functionality of AI agents and implement them in new areas, while simultaneously training employees to work with the new tools.

If this case sounds like what's happening in your company, our manager can help: he'll analyze your business and niche for free and point out where an AI agent would bring a real result in your case. Message the manager