AI applications have been built, deployed and used for industrial and government purposes for many years. The experiences have been documented in IAAI conference proceedings since 1989. Over that period, the breadth of applications has expanded many times over. The diversity of technical approaches has also evolved — from rule-based expert systems to deep learning — with many modern systems employing a variety of techniques and subsystems. This presentation will focus on contrasting what (we thought) we knew about building, deploying and using AI applications in the early years with what (we think) we know now.