Seems Somewhat Reasonable
So, whatever AI bullshit SAP is trying to peddle won't necessarily be as robust for on-prem users. I doubt that'll be any great loss, and some may see it as a feature because it means none of their data will be slurped up into this AI model, meaning no chance of it accidentally being leaked. It is at least possible that on-prem users could use their own data to train the AI. It may not be as "effective" given the smaller dataset, but it could be done.