Better AI doesn't remove complexity. It changes how you manage it.

That was the uncomfortable corollary running through WEM Live 2026 in Amsterdam—a day built around this argument: enterprise software is moving from code to models, and the organizations still treating AI as a faster way to write prompts are going to hit a wall the rest of the room has already found.

The shift goes through four recognizable stages. Vibe coding: describe what you want, get something probabilistic, buggy, and hard to maintain. Better AI code: more reliable, but still generated from prompts, and prompts don't scale. Language wasn't built to manage thousands of moving parts; describing a complex system in text is like trying to manage a city using paragraphs. The fourth stage is where this gets interesting: model-based creation, where you design visually, and the platform generates deterministic, production-ready code.

Prompts build features, models build systems: this distinction was the throughline of the day, and it's worth taking seriously regardless of whether you saw it argued live. 

CIOs, transformation leads, operations managers, data practitioners, and AI builders filled every seat in the TOBACCO Theater. Interestingly, the sessions that landed hardest weren't the ones selling the future; they were the ones reporting back from it. 

Enexis showed how it built its reuse operation on no-code, turning a recurring engineering problem into a repeatable system instead of a one-off fix. Ubbink walked through how real-time insight into warehouse processes turned a logistics blind spot into a live operational picture. And at Maastricht UMC+, the conversation wasn't about automation as an abstraction; it was process redesign in a hospital network, where "sustainable" means something has to hold up under regulatory scrutiny, not just survive a demo. 

CIOs, transformation leads, operations managers, data practitioners, and AI builders filled every seat in the TOBACCO Theater. Interestingly, the sessions that landed hardest weren't the ones selling the future; they were the ones reporting back from it. 

Enexis showed how it built its reuse operation on no-code, turning a recurring engineering problem into a repeatable system instead of a one-off fix. Ubbink walked through how real-time insight into warehouse processes turned a logistics blind spot into a live operational picture. And at Maastricht UMC+, the conversation wasn't about automation as an abstraction; it was process redesign in a hospital network, where "sustainable" means something has to hold up under regulatory scrutiny, not just survive a demo.