Jamie - This is interesting. In your opinion how did you expect call by name to work in the first place?
A speech recognizer is required detecting major languages and accents that will reliably catch "Snapp" and route it to the designated extension.
At least the way it is in 3CX right now, we just have a set of pre-recorded files to scan only. They are recorded in the user's actual voice so saved pitches and notations in the audio stream are denoted. Like this we route the call safer.