Intelligent Metallurgy: How AI Is Changing Processes
13 February 2026
From analytics to digital twins: AI transformation in steel manufacturing
Introduction: Why Metallurgy Needs Artificial Intelligence
Artificial intelligence in metallurgy is no longer an experiment or a buzzword. At The company, AI has become part of the production loop: it controls units, stabilizes technological regimes, reduces costs, and improves product quality. In her presentation, Svetlana Potapova, Head of Artificial Intelligence Cluster at a major steel manufacturing company, showed how the company progressed from early data experiments to comprehensive control systems and digital twins.
The key idea of the presentation is that AI's value is revealed only when it is embedded into the technological process and operates automatically, not merely as recommendations for humans.
Impact of AI Implementation: Numbers and Adoption of Solutions
One of the first questions when discussing AI in industry is the question of effect. At the company, it is measured not only by economic indicators but also by the actual use of solutions in production.
According to the company's accumulated practice:
- Increase in productivity of units and lines ranges from 3% to 11%, in some cases up to 15%.
-
Cost reduction: 1.5–5%.
-
Increase in product quality: about 30%.
-
Reduction in personnel labor costs due to NLP and GenAI solutions: 20–40%.
A separate emphasis is placed on the adoption rate: about 90% of deployed AI solutions are actually used. A total of 98 solutions are in industrial operation.
Evolution of AI at the company: From Data Lake to Digital Twins
The development of AI in the company occurred gradually. In the period 2017–2019 the main focus was on creating a Data Lake and accumulating data. This was a stage of experiments and searching for applicable scenarios.
In 2020–2021, advisory systems and analytical tools appeared, as well as the first video analytics projects. However, practice showed that recommendation systems do not provide a stable effect: personnel rarely use them continuously, and the solutions are quickly “forgotten.”
Starting in 2022, the company moved to comprehensive systems where ML, computer vision, and physics-mathematical models are directly integrated with the PCS (Process Control System) and automatically control units.
The target point of development is digital twins of shops, plants, and quarries, complemented by NLP and GenAI assistants to automate routine operations and increase labor productivity.