My organization has given preliminary approval for generative AI being used for initial transcription of audio, to then be verified by a human ear. That will likely be vastly more efficient. My unit has some fantastic civilian transcribers for our file disclosure. The general rule of thumb is one minute of audio is ten minutes of work- and that’s when done by properly equipped pros. It’s a great example of a task that can and should be done by violin support staff and not cops, although it’s fair to expect an officer to do a final review of their own audio against a transcript for court purposes.
It can be tough for us because we do large crown disclosures with potentially many dozens of statements, and often they want transcripts for all of them rather than summaries/will-says. New technology is gonna be great for this task. Hopefully we can start throwing AI at redaction/vetting as well.