You could set up a sidechain compressor with the dynamics processing. There is an effect in there to auto duck but bear in mind doing this in real time there will be some latency.
You’d need to set up the dynamics processing on the music track and side chain the voice track that’s armed for recording to the music train.
Remember though real time processing of effects in Audition may cause latency, it may be that she then finds this more off putting than not having the music while she records. Why not just have the monitor headphones volume a bit lower so she can hear the music bed without it being over powering and then do the auto-duck after the recording?