You throw audio into an LLM and get a function call back. Okay. An API call to OpenAI whisper costs around 0.01ct per minute, so really cheap. Feeding the text

You throw audio into an LLM and get a function call back. Okay. An API call to OpenAI whisper costs around 0.01ct per minute, so really cheap. Feeding the text-from-speech and the possible function calls with their JSON schema into the LLM costs another penny. Everything works. But now, where's your business model?