Idea of Integrating AI in robotics welding

In my company we have started to intergrate the use of UR-cobots to help us improve efficiency in our production, and later on ran into some issues regarding tracking of the welding seam with use of the Garmo - Garline tracker. The robot is a UR10e model and is equipped with a seam tracker (GARMO - Garline tracker). We find that there is an issue where it does not recognize the weld seam, causing it to lose track of the seam and deviate from the correct path. We could manually set the seam parametres, but since the object have different surfaces it becomes an challenge. I believe that if we integrate the AI camera, it will be able to detect the weld seam in advance and adjust the parameters for the Garline tracker accordingly, ensuring the correct parameters are applied when the weld seam varies. This is because the laser equipped on the Garline tracker is very sensitive to different materials, such as when the workpiece is rough, reflective, or inconsistent. I know this is maybe new, but a very interesting topic, since if this works it will change a lot of current problems.