Using interfaces to access the multitudes of information housed within a hotel's property management system (PMS), technology vendors have on their hands a literal treasure trove of first-party training data for machine learning applications. Any primary data extraction must nevertheless be focused around specific use cases that must be tested and verified before deploying in a live environment. Thus far, we've seen lots of discussion around what's possible in the abstract for machine learning, but outside of perhaps chatbots and algorithmic rate optimization recommendations, many of the purported use cases are not practical for the here and now.