KWS cannot replace free-form voice interaction; if users need to say commands like "set the living room light to 50%" that contain locations and values, the product typically still needs ASR and intent parsing.