If you have a homelab or vps you could try rolling your own with https://hub.docker.com/r/onerahmet/openai-whisper-asr-webservice for the transcription and a foss Tasker alternative for retreiving the voicemail when your dialler app posts a voicemail notification. If you don't have the capability to selfhost whisper, you'd probably have to sacrifice the illusion of privacy (your mobile carrier will always have access to any calls made over their network) and use a transcription service from google/microsoft/openai.