Like with all voice communications, context is important. There are only a limited number of things normally said and most transmissions are about things that were decided before the flight. Everyone involved is used to the bandwidth limited audio used here. Non-pilots have none of this context and normally have a hard time understanding these transmissions.

For ATC environments the voice data is a series of pre-expected prompts. You could do something different but you would pretty much have to redesign the whole system from scratch without making things significantly more complex. Complexity is the enemy of reliability.