AI Document Extraction behind the scenes, runtime performance, etc

Hi, I am evaluating Appian  OOTB capabilities for Document extraction. We have Trained an AI skill, used it in a process model, reconciled it, etc. But what is not written or documented anywhere is how is Appian doing this behind the scenes. I am trying to understand if a client is using Appian Cloud, where does the Model run does it use the same infrastructure, if yes then how does this affect the memory/cpu power (Models are generally cpu hungry is what i understand). So by introducing an AI skill to an existing model how does it affect the runtime performance of the process model, the memory consumption. Do we need to consider any upgrades in the Appian engine before using AI Skills.