No… basically these things need a trigger phrase, otherwise they’re continuously recording and running voice recognition and then NLP in the cloud for every sound in the environment, which is a lot of power investment (never mind privacy concerns) for minimal useful ROI.
The trigger phrase lets a simple, very low power dedicated chip with a simple model look for an audio clip that looks like the trigger phrase, which then kicks in recording and processing, usually off the local device, for whatever prompt follows.