You can encode voice with quite acceptable quality at 8kbps. (With recent advancements even 3kpbs is reasonably good.[1]) That’s 3.6MB per hour. Assuming an average of one hour per workday on the phone, a single employee generates less than 1GB a year. That’s peanuts. Even eight times that is lower than what I personally generate every day.

[1] https://ai.googleblog.com/2021/02/lyra-new-very-low-bitrate-...