In my team we've used the open source https://craig.chat/ and hooked it up to a Whisper API which lets us (1) control security of the recordings / transcripts; and (2) is ultra cheap if you sequence the transcription as bulk rather than real time.

Once we have the transcript there is then a pipeline which uses [insert your LLM of choice] and chains together prompts which (1) isolates only work-talk and then (2) produces various summaries for knowledge management / issue tracking.