AI meets BIM – taking Xella’s offer optimization to the next level

How exactly does AISOT work?

In short: engineers upload a 3D model, define rules and constraints, such as window and door positions, cutting limits, and the AI proposes an optimal solution. Broadly speaking, our goal is to minimize cutting intensity (i.e., saw meters), since additional cutting steps in production naturally increase delivery time. We work in a feedback loop: engineers review the suggestions, provide feedback, and the algorithm continuously adjusts its solutions. This creates a kind of "reinforcement learning", the AI can only perform certain actions, but it gets better step by step.

What effects have you seen so far?

AISOT has been in development since fall 2023. Through continuous optimization and a strong drive for change, we used the challenging construction years as an opportunity to unlock new efficiencies. As a result, we've achieved a significant productivity boost. We were able to nearly halve the effort required to generate quotations for our system wall elements. This means faster, more precise offers for our customers, and a major efficiency gain for us.

Who was involved?

The project is highly interdisciplinary. The AI incorporates the combined knowledge of our construction IT specialists and engineers. Together with external software developers, we tailored the AI to our specific needs. My role was to bring all areas together. My background in construction informatics and my experience at Xella made me an ideal interface. While Martin Müller, Head of Engineering & System Services, supported the drafting team's processes, Peter Lesage, Manager Group IT, contributed his expertise in IT development, prototyping, and applied AI. Aware that AI projects often venture into uncharted territory and carry high risk, he acted as a sounding board for the team.