Revived stock management thanks to AI

Published on 14/03/2024 in Innovate

Whether it is about streamlining finances or keeping customers happy, stock management is a crucial issue for any business in the retail sector. Proximus tackled the problem head-on thanks to the interaction between the AI and the logistics expert.

Revived stock management thanks to AI

Imagine having to manage several hundred references, product by product, manually. You would need to avoid stock-outs or surpluses, to manage busy periods such as sales or to target faulty products. All this would be done via an ERP system deployed on several screens. A nightmare, you might say! Well, that is how a specialist had to work for all Proximus shops every day. “Optimizing stocks was an art, almost an impossible task given the progressive increase in the number of references to be processed,” Jean-Christophe Goens, a specialist at Proximus, explains. But that was then!

AI is the backup of choice

Proximus decided to introduce AI as an additional layer of logic to optimize the existing ERP system. Wow, such a relief! “Technology has enabled us to increase the number of references we process. We’ve gone from 200 references to over 800,” says Bernard Löwenthal, Devices Team Manager at Proximus. Not to mention the considerable amount of time that we save. “In the new set-up, it only takes a few minutes a day for an operator”. The transition quickly enabled operators to concentrate on other tasks.

Technology has enabled us to increase the number of references we process. We’ve gone from 200 references to over 800

Bernard Löwenthal, Devices Team Manager at Proximus

An extremely precise system

The technology is based on a sophisticated Machine Learning model. “The accuracy requires a sufficient quantity of quality data, interaction with experts to constantly adjust the solution and a simple, pragmatic approach,” Guillaume Marion, Data Science Lead at Proximus ADA, explains. The data used in the ERP system in fact comes from a program, integrating information, technical data, prices and forecasts. “The actions are then carried out automatically by the algorithm, which intelligently combines the standard data. But they can also be manipulated according to the preferences of different logistics experts.”

Analytics Translator, a key role

One person was crucial in the initial phase of the new system: Aurélie, Proximus ADA’s Analytics Translator. Without her, the project would never have seen the light of day. Her role? “Collect the data, sort it out and understand it. She helped us to speak the same language as the AI. She was the link between us and Proximus ADA,” Goens explains. “Thanks to her, we understood what could and couldn’t be done with the data we had”. So, her role was essential in the initial phase of the project.

The benefits of the project are threefold: less excess stock, fewer stock-outs and an optimized workload

Bernard Löwenthal, Devices Team Manager at Proximus

A few telling numbers

“The benefits of the project are threefold: a reduction in surplus stock, a reduction in stock defects and an optimized workload,” Löwenthal stresses. By avoiding stock-outs and reducing surpluses, the financial impact has been significant with annual savings of several million euros. “It’s easy to explain: an out-of-stock situation costs an average of €150 in lost revenue per failed sale, while overstock represents a loss in value.” Eliminating these problems has meant savings for Proximus. “This has given us a 2% increase in ‘devices’ sales in the shops in 2023.”

A scalable solution

Given the success of the solution, Proximus wants to go further. “In the future, we plan to reduce our environmental footprint by optimizing the filling and frequency of mailboxes,” Goens continues. The collaboration with Proximus ADA is far from over. In particular, it will enable us to improve and further develop our algorithms for even more efficient stock management.


One magazine is the Proximus B2B magazine for CIOs and IT professionals in large and medium-sized organisations.

Other articles of One