In a lot of programs you have to enter text. When working on the road it is sometimes awkward to enter text. In these situations it may be easier to say something. Every text field should support recording a message. If possible it should try to convert the speech to text, but the speech should always be available until a person has entered text into the text field.
When the field needs to be understood by the computer (e.g. a phone number) the computer should require the human to convert the speech to text.
Activating voice text can be done by a gesture or maybe even by voice.
It should be possible to find all VoiceText-fields so you can convert all VoiceText fields if you have the time. It should be possible to export all VoiceText fields, send them to another system, do the conversion on another system, and import the converted fields without losing VoiceText fields that may have been made in the mean time. The conversion could be done remotely by a secretarial service. If done remotely it should be possible to check that the conversion was done correctly.