chevron-right Created with Sketch. back to thoughts

Streamlining procurement with Arcelor Mittal

Managing procurement in a multinational company might be challenging. Arcelor Mittal realized their procurement operations needed to be centralized. Therefore, they started searching for a solution to centralize their procurement operations over different entities of the company. Sagacify brought the company a big step closer to efficiency by building a customized AI solution for their problem.

Improve procurement with AI

Arcelor Mittal is a multinational company with divisions all over the world. Indeed, Arcelor Mittal noticed that different entities would order identical products separately from one another. This led to lost opportunities because if they had ordered these products together, they could have negotiated a better deal.

Finding the best solution

Sagacify solved this problem with its AI expertise. For this project, we applied NER (Named Entity Recognition) algorithms to extract and structure the manufacturers and ordered product references in the text descriptions. This allowed us to discover identical products and to deduplicate them. We also used active learning to optimize the labeling of the dataset, since manual labeling of the manufacturers and references was needed.

Obtaining ground truth data

Due to the lack of ground truth data, we created a labeled dataset through manual labeling.
In order to minimize the labeling work, we used Active Learning: a learning algorithm that is able to interactively query the users to obtain additional information about the data points that will maximize its learning rate. That way, our algorithm got a sufficient amount of labeled observations while keeping our client manual labeling work to a minimum.

Nevertheless, Sagacify managed to deliver positive results. The found references and duplicates were accurate and have shown useful for Arcelor Mittal to optimize its operations over different divisions. Still, there were some issues with one division because of the aforementioned data problem, but for another division, we managed to get 99% of the duplicates to be accepted. Nevertheless, close collaboration with the client was crucial for the success of this project and is something we greatly value here at Sagacify.

Do your operations have room for improvement? Come to us and we will look for the opportunities!

G e t i n t o u c h search Created with Sketch.