That's what I'm hoping to uncover ;p but if brainstorming about what could be interesting: - ability to alter tone/objective of the call based on the caller's intent (almost like a system-prompt rewrite) - contextual redaction of data - shorter delay (predictive buffering?) - continuous fine-tuning / learning feedback loop (with human-in-the-loop assessment) - A/B testing - custom chunking/indexing of RAG data (to be used by VAPI) - continue a disconnected call from where it was stopped (or maybe even call back the caller to continue conversation) - multimodal awareness (able to talk with the caller about details sent via text or email)