Nornickel Reduced Task Completion Time from 30 Days to Hours: How AI Agents Optimized Budget and Project Control

Nornickel, one of the world's largest metal producers, faced a challenge: given the company's scale, budget and project control processes were too time-consuming. The implementation of AI agents allowed for the reduction of task completion times from 30 days to just a few hours, significantly enhancing the efficiency of financial and project management.

In a large industrial company, where billions in budgets and hundreds of projects are managed, any delay in control is not just downtime; it's a direct financial loss. Manual data collection and analysis, approvals, endless re-checks—all this consumes time that could be spent on strategic planning and development. But today, this problem is solvable, and the solution lies in automating routine tasks with AI.

The Complexities of Budget and Project Management at Nornickel's Scale

Managing budgets and projects at a company like Nornickel involves a colossal amount of data, numerous stakeholders, and strict regulations. Daily operations require processing thousands of documents, monitoring the progress of hundreds of projects, comparing actual expenses with planned figures, and generating reports for various management levels. All of this demanded significant human resources and time.

Traditional control methods, based on manual data collection and analysis, led to lengthy approval cycles and decision-making processes. For example, preparing a comprehensive report or analyzing deviations for a large project could take weeks. This slowed down reactions to market changes, complicated operational management, and increased risks.

Why Standard Solutions Were Insufficient

The company already used various information systems for financial and project management. However, most of them were focused on structured data and performing clearly defined operations. As soon as there was a need to analyze unstructured information, correlate data from different systems, or generate non-standard reports, human intervention was required. These systems were good for accounting but not for deep analysis and real-time decision support.

This is why Nornickel turned to the concept of AI agents – systems capable of not just processing information according to a given algorithm, but also understanding context, extracting meaning from natural language, and autonomously performing complex tasks requiring analytical capabilities.

How AI Agents Were Designed for Nornickel

The solution is based on a hybrid architecture that combines Nornickel's own computing power and cloud infrastructure. This allowed for the creation of a secure and scalable environment for working with large language models. Nornickel became the first company in the metallurgical sector to use its own large language models (LLMs) within its corporate network, without data leaving the internal perimeter, which is critical for security.

A proprietary MetalGPT model was developed for creating AI agents. The agents were designed as digital assistants capable of performing the following functions:

• Budget Control. Automatic collection and analysis of expense and revenue data from various sources, identification of budget deviations, forecasting, and recommendations.
• Project Control. Monitoring project progress, risk analysis, identification of bottlenecks, and proposals for optimizing timelines and resources.
• Inventory Management. Optimizing inventory levels, forecasting material needs, and reducing storage costs.
• Financial Analysis. Preparing analytical reports, identifying trends, and supporting investment and financial planning decisions.

A key requirement was the agents' ability to understand natural language queries and provide meaningful answers and recommendations, rather than just raw data.

Implementation and Scaling

The implementation of AI agents was carried out in stages, starting with the most critical and labor-intensive processes. Tasks where the effect of reducing lead times was most tangible were automated first. Simultaneously, employees were trained to effectively interact with the new digital assistants. It was important not just to provide a tool, but also to teach the team how to maximize its potential.

The next step was to expand access to AI tools for a wider range of employees, including through a specialized module for creating simple AI agents without deep programming skills. This approach not only allowed for the implementation of ready-made solutions but also stimulated internal development of new AI agents for specific departmental needs.

Results

Reducing task completion times from 30 days to several hours allowed Nornickel to significantly accelerate decision-making processes, enhance operational responsiveness to changes, and mitigate risks. This directly impacts the company's economic efficiency, enabling more flexible budget and project management.

How to Replicate This in Your Business

Nornickel's case demonstrates that implementing AI agents can bring significant economic benefits even in complex industrial sectors. If your company faces similar challenges, here's where you can start:

• Identify "pain points." Find processes that consume a lot of time, require manual labor, or slow down decision-making. This could include data collection for reports, analysis of large volumes of information, or routine approvals.
• Assess security. For handling confidential data, consider deploying LLMs within your own perimeter or using hybrid solutions, as Nornickel did, to ensure maximum security.
• Start small, scale gradually. Choose one or two pilot scenarios where the impact will be most noticeable. After successful implementation and demonstration of results, scale the solution to other processes.
• Engage employees. Train your team to work with AI agents and encourage them to find new application scenarios. This will help unlock the technology's full potential.

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