When DHL decided to incorporate artificial intelligence into its voice assistant, it did so with a very different approach from many other companies. The project began with a simple problem. The system was unable to reliably recognize the German “ja,” the equivalent of “yes.” That misstep revealed that, before scaling, it was essential to fix the basics. The company treated AI as part of its critical infrastructure rather than rushing into eye-catching innovation projects.

DHL understood that speed without solid foundations leads to failure, and that governance is part of progress rather than an obstacle to it. For six months, teams from different departments worked in an isolated lab. They agreed on clear rules and cleaned up stored data without interfering with live operations. The result was a voice assistant that now handles around one million calls each month and resolves half of them without human intervention. The case shows that it is often more effective to create suitable environments for research and testing, and only then scale up sustainably. Truly novel tools do not always come from hurried, large-scale investments. Real progress lies in integrating these technologies into processes validated by practice, supported by reliable data and clear budgets.