The presentation follows the integration of an ML solution (SmartNav, for navigation or SmartDrive for vehicle handling), using the Machine Learning Framework (ML.F) in a Ubisoft engine (Snowdrop and Anvil) from start to end. It will focus on the complete pipeline use to enable ML in games, especially the costs (like resource, time, CPU/RAM, etc.) and the engine requirements. It will also present the potential of this new tool, while also exposing its cons to help people know if ML is right for the job. And finally, we present how sharing knowledge among many ML-based projects can help.